15.10.02 · orgchem / retrosynthesis

Multi-step synthesis design: protecting groups, convergent versus linear strategies

stub3 tiersLean: nonepending prereqs

Anchor (Master): Carey & Sundberg — Advanced Organic Chemistry, Part B, Ch. 1

Intuition Beginner

Building a complex molecule from simple starting materials is like assembling a machine from parts. You cannot always install every part at once — some components interfere with each other during assembly. In organic synthesis, the solution is the protecting group: a temporary mask that covers a reactive functional group so it does not get damaged while you work on another part of the molecule.

Retrosynthetic analysis works backward from the target. You break bonds one at a time until every piece is a commercially available starting material. Once the plan is complete, you run the reactions forward in the laboratory. Each forward step must be compatible with every functional group present at that stage — and when it is not, a protecting group bridges the gap.

There are two ways to organise a multi-step plan. Linear synthesis builds the molecule in a single chain: A to B to C to the target. Convergent synthesis builds two large fragments separately and joins them in one final step. Convergent plans generally give higher overall yield because each fragment passes through fewer steps, losing less material along the way.

Visual Beginner

Picture a target molecule as a finished building. Retrosynthetic analysis tears the building down to its prefabricated modules. Protecting groups are the scaffolding that holds one wing in place while construction proceeds on another.

The convergent plan has two branches of three steps each plus one joining step. The longest linear sequence is four steps. The linear plan has seven steps. Even if each step has the same yield, the convergent plan preserves more material because no single molecule travels through more than four steps.

Worked example Beginner

Target: methyl 4-hydroxybutanoate from 1,4-butanediol.

The target contains an ester group and a free alcohol. To form the ester, you need to oxidise one of the two hydroxyl groups in 1,4-butanediol to a carboxylic acid and then esterify it with methanol. But a typical oxidant such as chromic acid would oxidise both hydroxyl groups, destroying the one you intend to keep.

Step 1. Protect one hydroxyl group as a TBDMS ether using TBDMSCl and imidazole. This masks one $- O H$ as an inert silyl ether.

Step 2. Oxidise the remaining free hydroxyl to a carboxylic acid using Jones reagent ( $C r O_{3} / H_{2} S O_{4}$ ). The TBDMS-protected alcohol is unaffected.

Step 3. Esterify the carboxylic acid with methanol and acid catalysis to give the methyl ester.

Step 4. Remove the TBDMS protecting group with TBAF (tetrabutylammonium fluoride) to reveal the free alcohol. The product is methyl 4-hydroxybutanoate.

The protecting group was necessary because the oxidation step would have affected both hydroxyl groups without it.

Check your understanding Beginner

Formal definition Intermediate+

A multi-step synthesis is a sequence of chemical reactions that transforms commercially available starting materials into a target molecule through a series of intermediates. The design of a multi-step synthesis involves three interdependent decisions: the disconnection strategy (which bonds to form and in what order), the protecting-group strategy (which functional groups to mask and when), and the topology (linear versus convergent arrangement of steps).

Protecting-group orthogonality

Two protecting groups are orthogonal if the conditions for removing one do not affect the other. Orthogonality allows selective deprotection in multi-functional targets. The major protecting-group families and their orthogonality classes are:

Functional group	Protecting group	Abbreviation	Installation	Removal	Orthogonality class
Alcohol	tert-Butyldimethylsilyl ether	TBDMS	TBDMSCl, imidazole	TBAF or acid	Fluoride-labile
Alcohol	Methoxymethyl ether	MOM	MOMCl, base	Acid ( $H C l$ )	Acid-labile
Alcohol	Acetate ester	Ac	Acetic anhydride, pyridine	Base hydrolysis ( $K_{2} C O_{3}$ )	Base-labile
Alcohol	Benzoyl ester	Bz	Benzoyl chloride, pyridine	Base hydrolysis ( $N a O H$ )	Base-labile
Amine	tert-Butoxycarbonyl	Boc	$(B oc)_{2} O$ , base	TFA or HCl	Acid-labile
Amine	Benzyloxycarbonyl	Cbz	CbzCl, base	Hydrogenolysis ( $H_{2} / P d$ )	Reductive
Amine	9-Fluorenylmethoxycarbonyl	Fmoc	FmocCl, base	Piperidine (base)	Base-labile
Aldehyde/Ketone	Acetal	—	Diol or dithiol, acid	Aqueous acid	Acid-labile
Carboxylic acid	Methyl ester	Me	$C H_{2} N_{2}$ or MeOH, acid	LiOH or NaOH	Base-labile

An orthogonal set for a molecule containing an amine, an alcohol, and a carboxylic acid might use Boc (acid-labile) for the amine, TBDMS (fluoride-labile) for the alcohol, and a methyl ester (base-labile) for the acid. The three removal conditions — acid, fluoride, and base — are mutually compatible.

Convergent versus linear step count analysis

For a linear plan of $n$ steps at per-step yield $y$ , the overall yield is $Y_{L} = y^{n}$ .

For a convergent plan with two branches of $p$ and $q$ steps joined by one coupling step at yield $y_{j}$ , the overall yield is $Y_{C} = y^{p} \cdot y^{q} \cdot y_{j}$ .

The longest linear sequence (LLS) is the maximum number of sequential steps any single molecule passes through. For the linear plan, LLS $= n$ . For the convergent plan, LLS $= max (p, q) + 1$ . Since the material throughput scales as $y^{LLS}$ , the convergent plan is preferred whenever it reduces the LLS and the joining step yield $y_{j}$ is at least as high as the average per-step yield $y$ .

Functional group interconversion table

When a direct disconnection is not available, the planner converts one functional group into another that is easier to disconnect. Common functional group interconversions (FGIs):

Starting group	Target group	Reagent/conditions	Purpose in retrosynthesis
Ketone ( $C = O$ )	Alkene ( $C = C$ )	Wittig or Tebbe reagent	Changes disconnection from aldol-type to olefination
Alcohol ( $- O H$ )	Alkene ( $C = C$ )	Dehydration ( $H_{2} S O_{4}$ )	Enables elimination-based disconnections
Alkene ( $C = C$ )	Alcohol ( $- O H$ )	Hydroboration-oxidation	Installs alcohol for later protection or activation
Alkene ( $C = C$ )	Diol ( $- O H$ , $- O H$ )	$O s O_{4}$ , then $N a H S O_{3}$	Provides 1,2-diol for acetonide protection
Aldehyde ( $- C H O$ )	Alcohol ( $- O H$ )	$N a B H_{4}$	Reduces reactivity for selective manipulations
Alcohol ( $- O H$ )	Aldehyde ( $- C H O$ )	$P C C$ or Swern oxidation	Activates for nucleophilic addition
Nitrile ( $- C N$ )	Carboxylic acid ( $- C O O H$ )	$H_{3} O^{+}$ , heat	Extends carbon chain by one atom

Umpolung: the d1 synthon concept

In normal polar disconnections, a carbonyl carbon acts as an electrophilic acceptor (a1 synthon). Umpolung reverses this polarity, making the carbonyl carbon act as a nucleophilic donor (d1 synthon). The most important umpolung reagent is the 1,3-dithiane: treatment of an aldehyde with 1,3-propanedithiol under acid catalysis gives a dithiane, whose C-2 carbon is deprotonated by $n$ -BuLi to give a nucleophilic carbanion. This d1 synthon reacts with alkyl halides and carbonyl compounds, expanding the retrosynthetic search space by allowing disconnections whose polarity does not match the native functional-group reactivity.

Counterexamples to common slips

Using the same protecting group for two identical functional groups prevents selective deprotection of one but not the other. If a molecule has two primary alcohols that must be deprotected at different stages, they require different protecting groups from orthogonal classes.
Assuming convergent synthesis always beats linear synthesis. If the joining step has low yield (e.g., a macrocyclisation at 15%), the convergent plan may give lower overall yield than a linear alternative. The convergent yield proposition $Y_{C} > Y_{L} ⟺ y_{j} > y$ must hold.
Forgetting the protecting-group tax. Each protecting group adds two steps (installation and removal). Four protecting groups add eight steps, costing material even if each protection/deprotection is 95% yield ( $0.9 5^{8} = 66%$ of material consumed before productive chemistry).

Key theorem with proof Intermediate+

Proposition (Optimal branch-point yield). Consider a convergent plan with $k$ branch points, where branch $i$ has $n_{i}$ steps at per-step yield $y$ and the $i$ -th coupling step has yield $y_{i}$ . The overall yield is

Y_{C} = (i = 1 \prod k y^{n_{i}}) \cdot (i = 1 \prod k y_{i}) = y^{N_{branch}} \cdot i = 1 \prod k y_{i}

where $N_{branch} = \sum_{i = 1}^{k} n_{i}$ is the total number of non-coupling steps across all branches. If the equivalent linear plan has $N_{branch} + k$ steps, then the convergent advantage is

\frac{Y _{C}}{Y _{L}} = \frac{y ^{N_{branch}} \cdot \prod _{i = 1}^{k} y _{i}}{y ^{N_{branch} + k}} = i = 1 \prod k \frac{y _{i}}{y} .

The convergent plan is superior if and only if the geometric mean of the coupling-step yields exceeds the per-step yield of the linear branches.

Proof. The linear plan yield is $Y_{L} = y^{N_{branch} + k} = y^{N_{branch}} \cdot y^{k}$ . The convergent plan yield is $Y_{C} = y^{N_{branch}} \cdot \prod_{i = 1}^{k} y_{i}$ . Dividing:

\frac{Y _{C}}{Y _{L}} = \frac{y ^{N_{branch}} \cdot \prod _{i = 1}^{k} y _{i}}{y ^{N_{branch}} \cdot y ^{k}} = \frac{\prod _{i = 1}^{k} y _{i}}{y ^{k}} = i = 1 \prod k \frac{y _{i}}{y} .

Since each factor $y_{i} / y > 1$ contributes a multiplicative advantage, the convergent plan is superior precisely when $\prod_{i = 1}^{k} (y_{i} / y) > 1$ , which holds if and only if $(\prod_{i = 1}^{k} y_{i})^{1/ k} > y$ . $□$

This result generalises the single-branch-point proposition from 15.10.01 to multi-branch convergent plans and quantifies the cumulative advantage of each high-yielding coupling step.

Exercises Intermediate+

Exercise 4 (medium, numeric).

A linear 10-step synthesis has per-step yield of 80%. A convergent alternative uses two 4-step branches at 80% per step, joined by a Suzuki coupling at 95% yield, followed by one final step at 80% yield. Compute both overall yields as percentages rounded to one decimal place.

Hint

Linear: $0.8 0^{10}$ . Convergent: $0.8 0^{4} \times 0.8 0^{4} \times 0.95 \times 0.80$ . The convergent plan has $4 + 4 + 1 + 1 = 10$ total steps but the LLS is only $4 + 1 + 1 = 6$ .

Answer

Linear: $0.8 0^{10} = 0.1074 = 10.7%$ .

Convergent: each branch gives $0.8 0^{4} = 0.4096$ . After coupling at 95%: $0.4096 \times 0.4096 \times 0.95 = 0.1594$ . Final step at 80%: $0.1594 \times 0.80 = 0.1275 = 12.8%$ .

The convergent plan gives 12.8% versus 10.7% for the linear plan. The advantage comes from reducing the longest linear sequence from 10 to 6 steps, despite having the same total step count.

Exercise 5 (medium, short-answer).

Explain the concept of umpolung using the 1,3-dithiane strategy. Why does it expand the retrosynthetic search space?

Hint

Normally a carbonyl carbon is electrophilic. What happens when you convert the carbonyl to a dithiane and deprotonate?

Answer

In normal reactivity, a carbonyl carbon is electrophilic (acceptor, a1 synthon). The 1,3-dithiane strategy converts the aldehyde to a dithiane, then deprotonates at C-2 with $n$ -BuLi to give a nucleophilic carbanion (donor, d1 synthon). This polarity inversion is umpolung. It expands the search space because the planner can now disconnect bonds adjacent to the carbonyl in either polarity direction — the carbonyl carbon can serve as either nucleophile or electrophile — effectively doubling the number of viable disconnections at that site.

Exercise 6 (hard, short-answer).

A target molecule requires five protecting groups (each installed and removed at 95% yield). Calculate the protecting-group tax as a fraction of material consumed. Then propose two strategies for reducing this tax without changing the target.

Hint

Five protecting groups means ten extra steps (five installations, five removals). Compute $0.9 5^{10}$ . Then think about whether all five are truly necessary or whether some can serve double duty.

Answer

Protecting-group tax: $0.9 5^{10} = 0.5987$ . Approximately 60% of material is consumed by protection and deprotection alone, before any productive chemistry occurs.

Strategies to reduce the tax: (1) Eliminate protecting groups by using chemoselective reagents that transform only the desired functional group — Baran's protecting-group-free philosophy. (2) Use protecting groups that serve double duty, such as a TBDPS group that both protects an alcohol and acts as a chromatographic handle for purification. (3) Redesign the route so that a functional group is installed after the interfering step rather than protected before it, avoiding the need for a protection/deprotection pair entirely.

Exercise 7 (hard, short-answer).

Design a convergent retrosynthetic analysis for the following target: a molecule with two aromatic rings connected by a three-carbon chain containing a ketone at the central carbon. One ring has a hydroxyl group (phenol); the other has a primary amine. Propose protecting groups, identify the branch point, and justify the convergent topology.

Hint

The ketone in the three-carbon linker is a natural branch point — disconnect adjacent to it. The phenol and amine may need protection during the coupling step.

Answer

Branch point: disconnect at the C-C bond between the ketone carbon and the chain carbon adjacent to ring A. This gives two synthons — an enolate donor derived from a ring-A-substituted fragment and an electrophilic acceptor derived from a ring-B-substituted fragment.

Branch A: the phenolic ring with a two-carbon enolate precursor. Protect the phenol as a MOM ether (acid-labile) during the coupling step. The enolate synthetic equivalent is the corresponding methyl ketone treated with LDA.

Branch B: the amine-bearing ring with a one-carbon electrophilic fragment. Protect the amine as Boc (acid-labile) during the coupling step. The electrophilic synthetic equivalent is the corresponding acyl chloride or Weinreb amide.

Coupling: enolate addition to the electrophile gives the ketone. Deprotect MOM with mild acid and Boc with TFA in a single step (both are acid-labile and can be removed simultaneously). The convergent topology is justified because each branch is synthesised independently — the phenol and amine never interfere with each other's branch chemistry — and the coupling step is a high-yielding carbonyl addition.

Protecting-group economy and cascade reactions Master

The most elegant multi-step syntheses minimise the number of discrete operations — including protecting-group installations and removals — by designing reaction sequences where each step's product is the direct substrate for the next without intermediate purification or protection. Two advanced strategies achieve this: protecting-group economy and one-pot cascade reactions.

Protecting-group economy

Protecting-group economy measures the ratio of productive bond-forming steps to protecting-group steps in a synthesis. A synthesis with 12 productive steps and 8 protecting-group steps has a protecting-group economy of $12/20 = 60%$ , meaning 40% of the total operations are devoted to masking and unmasking functionality. Baran's synthesis of haouamine A achieved a protecting-group economy of 100% — zero protecting-group steps — by selecting reagents whose chemoselectivity matched the target's functional-group hierarchy.

The principle of protecting-group economy states: protecting groups should be used only when no chemoselective alternative exists. The planner evaluates each proposed protecting-group step against three criteria:

Chemoselectivity test. Can the reaction be performed without the protecting group by choosing a more selective reagent? If so, the protecting group is unnecessary.
Sequential installation test. Can the functional group be installed after the interfering step instead of protected before it? If so, the protecting group is unnecessary.
Convergent topology test. Can the synthesis be rearranged so that the interfering functional groups are on different convergent branches, never meeting until the final coupling? If so, protecting groups on one branch are unnecessary during the other branch's synthesis.

When all three tests fail, the protecting group is genuinely necessary and the planner selects it from the orthogonal set that maximises compatibility with the rest of the route.

One-pot cascade reactions

A cascade reaction (also called a domino or tandem reaction) is a sequence of two or more transformations that occur in a single reaction vessel without isolating intermediates. Each transformation generates the substrate for the next automatically. Cascade reactions eliminate the protecting-group tax because no intermediate is isolated and no protecting group is needed to bridge between steps.

Corey's prostaglandin synthesis uses a cascade strategy: a Diels-Alder reaction constructs the cyclopentene ring, and the resulting product spontaneously undergoes an intramolecular lactonisation without any intermediate manipulation. The cascade replaces what would otherwise require three separate steps (Diels-Alder, protection, lactonisation) with one operation.

The planning principle for cascade reactions is: identify sequences of transforms where the product of transform $n$ is the direct substrate for transform $n + 1$ . The retrosynthetic planner looks for patterns in the target where a bond formed in one disconnection creates the structural motif required for the next disconnection. The Robinson annulation is a classic cascade: a Michael addition followed by an intramolecular aldol condensation, both occurring in the same pot.

Total synthesis planning philosophy

The design of a total synthesis — the laboratory preparation of a complex natural product from simple starting materials — integrates all of the principles above into a single coherent plan. Corey articulated the philosophy as follows ^{[Corey and Cheng 1989]}:

Simplicity. Prefer the shortest route with the fewest steps. Each step is a potential failure point.
Convergence. Prefer convergent topologies that reduce the longest linear sequence.
Selectivity. Prefer reagents and conditions that are chemo-, regio-, and stereoselective, avoiding the need for protecting groups.
Reliability. Prefer well-established reactions with broad substrate tolerance over novel reactions with uncertain scope.
Flexibility. Design the route so that if one step fails, an alternative is available without redesigning the entire plan.

These five principles are sometimes in tension. A novel reaction may offer superior step economy (principle 1) but lower reliability (principle 4). A convergent topology (principle 2) may require more protecting groups (conflicting with principle 3). The planner's judgement lies in weighing these trade-offs for each specific target.

Computer-assisted synthesis planning Master

Corey's LHASA program, introduced in the late 1960s, was the first computational tool for retrosynthetic analysis ^{[Corey and Cheng 1989]}. LHASA encoded several hundred named reactions as transforms — each with a structural recognition pattern, a disconnection rule, and scope-limiting filters — and applied them recursively to generate synthon trees. The chemist guided the search through a graphical interface, selecting which branches to explore.

Modern tools extend LHASA's approach with machine learning:

Synthia (Chematica) ^{[Grzybowski 2018]} constructs a network graph of all known reactions and searches for paths from the target to commercially available starting materials, scoring routes by predicted yield, selectivity, and starting-material cost.
ASKCOS (MIT) and AiZynthFinder (AstraZeneca) use graph neural networks to predict which bond to disconnect and which transform to apply, then recursively apply disconnections until starting materials are reached.
Retro and Retro+,** developed at Berkeley, use Monte Carlo tree search guided by neural-network policies to explore the synthon tree efficiently.

These tools are most effective for targets with 10-20 heavy atoms and moderate stereochemical complexity. For larger targets, the combinatorial explosion of possible disconnections and the difficulty of predicting stereochemical outcomes limit their reliability. The gap between computational prediction and experimental reality remains nontrivial: roughly 70% of proposed routes for moderately complex targets execute as planned, while the remaining 30% require human intervention to troubleshoot unexpected selectivity problems or reagent incompatibilities.

The philosophical significance is that retrosynthetic planning is a form of reasoning under uncertainty. The planner makes decisions based on predicted outcomes and revises as experimental data arrives. Computer-assisted tools augment this reasoning by expanding the search space beyond what a human planner can hold in memory, but the feedback loop between planning and execution remains the defining methodological structure of synthetic chemistry.

Connections Master

Retrosynthetic analysis 15.10.01. This unit applies the disconnection framework established in the prerequisite unit to the practical design of multi-step routes. Protecting groups and convergent topology are the two principal tools for managing the complexity of real synthetic plans.
Carbonyl chemistry — nucleophilic addition 15.07.01. The majority of productive bond-forming steps in multi-step synthesis involve carbonyl addition reactions (Grignard, aldol, Michael, Wittig). Protecting groups are often needed precisely because carbonyl reagents (organolithiums, Grignards) are incompatible with other carbonyl groups in the molecule.
Amino acids and protein chemistry 15.12.01. Solid-phase peptide synthesis applies convergent and linear synthesis principles at every amide bond. The Fmoc/ $t$ -Bu orthogonal protecting-group set is the standard for peptide chain assembly, and the protecting-group economy of peptide synthesis is a major concern for long sequences.
Nucleic acid chemistry 15.13.01. Oligonucleotide synthesis is a linear multi-step plan analogous to peptide synthesis, with phosphodiester bond formation as the iterative coupling step. Convergent fragment coupling extends this to longer sequences.
Chemical kinetics 14.08.01. The yield of each step in a multi-step synthesis is determined by its kinetics. The planner uses kinetic data from model reactions to estimate per-step yields and to select conditions that maximise selectivity.
Electrophilic addition to alkenes 15.05.01. Functional-group interconversions between alkenes and alcohols (hydroboration-oxidation, oxymercuration-reduction) are frequently used in multi-step synthesis to install or remove hydroxyl groups at specific positions.

Historical and philosophical context Master

The concept of protecting groups emerged organically from the practice of carbohydrate chemistry in the late 19th and early 20th centuries. Emil Fischer's synthesis of glycosides required selective manipulation of individual hydroxyl groups on sugar molecules, and acetylation and benzylation were used as protecting strategies long before the term "protecting group" was formalised.

The systematic study of protecting-group orthogonality accelerated with the development of peptide synthesis in the 1950s and 1960s. Bruce Merrifield's solid-phase peptide synthesis (Nobel Prize, 1984) required an orthogonal protecting-group pair that could be removed selectively during iterative chain elongation. The Boc/benzyl and later Fmoc/ $t$ -Bu orthogonal sets were developed to meet this requirement, and their success established orthogonal protection as a general principle of multi-step synthesis design.

The convergent versus linear distinction was formalised by Corey and coworkers in the context of total synthesis planning in the 1960s and 1970s. Corey's synthesis of longifolene (1964) demonstrated convergent planning, and the quantitative yield analysis comparing convergent and linear topologies became a standard part of retrosynthetic pedagogy through Warren's Organic Synthesis: The Disconnection Approach ^{[Warren 2008]}.

The protecting-group-free synthesis movement, championed by Phil Baran beginning in 2007, represents a philosophical counterpoint. Baran argued that over-reliance on protecting groups reflects a failure of creativity in reagent selection, and that the most elegant syntheses exploit innate chemoselectivity rather than masking functionality. The debate between protecting-group-rich and protecting-group-free approaches continues, with most practitioners recognising that both strategies have their place depending on the target.

Bibliography Master

@book{McMurry2019,
  author = {McMurry, J.},
  title = {Organic Chemistry},
  edition = {10th},
  publisher = {Cengage},
  year = {2019}
}

@book{Clayden2012,
  author = {Clayden, J. and Greeves, N. and Warren, S.},
  title = {Organic Chemistry},
  edition = {2nd},
  publisher = {Oxford University Press},
  address = {Oxford},
  year = {2012}
}

@book{CareySundberg2007,
  author = {Carey, F. A. and Sundberg, R. J.},
  title = {Advanced Organic Chemistry, Part {B}: Reaction and Synthesis},
  edition = {5th},
  publisher = {Springer},
  address = {New York},
  year = {2007}
}

@book{CoreyCheng1989,
  author = {Corey, E. J. and Cheng, X.},
  title = {The Logic of Chemical Synthesis},
  publisher = {Wiley},
  address = {New York},
  year = {1989}
}

@book{Warren2008,
  author = {Warren, S.},
  title = {Organic Synthesis: The Disconnection Approach},
  edition = {2nd},
  publisher = {Wiley},
  address = {Chichester},
  year = {2008}
}

@article{Seebach1979,
  author = {Seebach, D.},
  title = {Methods of Reactivity {Umpolung}},
  journal = {Angew. Chem. Int. Ed. Engl.},
  volume = {18},
  pages = {239--258},
  year = {1979}
}

@article{Grzybowski2018,
  author = {Klucznik, T. and Mikulak-Klucznik, B. and McCormack, M. P. and Lima, H. and Szymku\'{c}, S. and Bhowmick, M. and Molga, K. and Zhou, Y. and Rickershaw, L. and Gajewska, E. P. and Toutov, A. and Dixo, J. M. and Molga, K. and Grzybowski, B. A.},
  title = {Efficient Syntheses of Diverse, Medicinally Relevant Targets Planned by Computer and Executed in the Laboratory},
  journal = {Chem},
  volume = {4},
  pages = {522--532},
  year = {2018}
}

@book{NicolaouSorensen1996,
  author = {Nicolaou, K. C. and Sorensen, E. J.},
  title = {Classics in Total Synthesis},
  publisher = {VCH},
  address = {Weinheim},
  year = {1996}
}