By mass, an Fmoc solid-phase peptide synthesis is mostly solvent. Coupling, Fmoc removal and the washes between them all run in the same bulk polar aprotic medium, and one of the main environmental problems with the method is the quantity of that solvent and the waste it becomes [1]. The scale is not marginal. When fourteen member companies of the ACS Green Chemistry Institute Pharmaceutical Roundtable pooled process mass intensity (PMI) data across 40 synthetic peptide processes at various development stages, SPPS averaged a PMI of approximately 13,000, set against a median of 168–308 for small molecules and approximately 8,300 for biopharmaceuticals [2].
What forced the issue was not the mass but the classification. N,N-dimethylformamide is considered reprotoxic, and its replacement is framed in the primary literature as urgent rather than aspirational [1]. On 12 December 2023 the European Commission amended Annex XVII of REACH to restrict DMF on the EU market on toxicity grounds [3]. N-methylpyrrolidone is named alongside DMF as a hazardous SPPS solvent generating substantial waste [4], so the obvious first move — swap one dipolar aprotic amide for another — solves nothing.
What DMF is doing that a replacement has to do
Two jobs, and they are not the same job. The solvent has to swell the resin so that reagents reach the growing chain inside the bead, and it has to dissolve everything you put in: protected amino acids, coupling reagents, additives and the by-products released during the cycle. A candidate is therefore screened across all steps — amino acid solubility, resin swelling, deprotection kinetics and coupling — before it goes anywhere near a real target [5]. There is a third requirement that is easy to forget: the replacement must not favour side reactions more than DMF does [1].
Swelling is the property measured first and misread most often. A systematic study of polystyrene- and PEG-based resins in greener solvents found that chemical factors — the nature of the functionality or linker, and the degree of loading — had a far larger influence on swelling than physical parameters such as bead size [6]. Interpreted through Hansen solubility parameters, the data indicated that some solvents interact with the polymeric core of the resin while others interact with the functionality [6]. The same work extended the measurements to the changes in swelling that occur during deprotection and chain elongation [6], which is the practically important part: the number you measure on a bare resin is not the swelling you have at residue twenty.
Binary mixtures do not interpolate
If no single green solvent matches DMF, the obvious move is to mix two. That works, but not in the way a linear-blending intuition suggests. Swelling of Merrifield and HypoGel 200 resins in two-solvent mixtures was generally found not to vary linearly with the relative amount of each solvent [7]. Mapping mixtures into Hansen solubility parameter space defined high, medium and low swelling regions for each resin, and the swelling of a given binary mixture could then be explained by how the line joining the two pure solvents cut through that resin's swelling region [7]. The consequence is concrete: an appropriately chosen mixture of two green solvents was more effective for solid-phase peptide synthesis on Merrifield resin than either individual solvent, and could completely replace the traditional polar aprotic and chlorinated solvents for that application [7].
A parallel line of work approached mixture design from bulk physical properties rather than solubility parameters, reporting that binary mixtures with a suitable polarity and viscosity profile are viable green alternatives to DMF in SPPS [8]. A follow-up study varied mixture composition systematically and tracked its effect on Fmoc removal, coupling and the common side reactions, finding that the purity profile was not adversely affected relative to DMF [9]. The more interesting result was that composition becomes a control variable within a synthesis: varying the solvent ratio at a single step significantly suppressed a problematic Arg-lactamisation in a 7.5 mmol synthesis of the 20-mer bivalirudin, and composition changes were also used to suppress aspartimide formation [9]. That handle does not exist when you run neat DMF.
Coupling is easy; Fmoc removal is the constraint
The most useful mechanistic statement in this literature is that the two reactions of the cycle have different solvent requirements. Coupling proceeds readily in solvents of significantly lower polarity than DMF, but high polarity is essential for Fmoc removal using piperidine, and it is that asymmetry which limits the green solvent options [10]. Once the constraint is located on the deprotection, you can attack the base instead of the solvent. Pyrrolidine was identified as an efficient base for Fmoc removal in less polar binary mixtures that also favour coupling — dimethyl sulfoxide/ethyl acetate 1:9, and N-butylpyrrolidinone (NBP)/1,3-dioxolane at 2:8 and 4:6 [10]. Crude purities from those less polar mixtures with pyrrolidine were comparable to or better than those from DMF with piperidine on challenging targets, and scaled-up syntheses at 5 and 7.5 mmol respectively of the 29-mer dasiglucagon and the 20-mer bivalirudin gave purity profiles amenable to further optimisation [10].
The caveat sits in the same paper: evaluation of base-dependent side reactions showed increased side-product formation with pyrrolidine on diketopiperazine- and aspartimide-prone sequences [10]. This is the recurring shape of the field. Nothing here is a free swap; each substitution moves the side-reaction profile rather than removing it, and which direction it moves depends on the sequence.
Deprotection kinetics should be measured rather than assumed. In dipropyleneglycol dimethylether (DMM) with 20% piperidine, the Fmoc-Val-OH signal disappeared after 30 minutes, against 6 minutes for the same measurement in DMF [5]. A five-fold slower removal is not disqualifying, but it changes cycle time and it changes how long a base-sensitive sequence sits in piperidine.
The solvent can consume your reagent
γ-Valerolactone (GVL) is a frequently proposed green SPPS solvent with a specific failure mode. GVL is susceptible to ring opening in the presence of bases such as the piperidines used for Fmoc removal; time-dependent NMR monitoring, corroborated by theoretical calculations, showed formation of the corresponding acyl piperidides, and the authors concluded that a solution of piperidines in GVL should be prepared daily for better Fmoc removal [11]. Nothing in a swelling table or a solubility screen predicts this. If you are qualifying a candidate solvent, the deprotection cocktail needs its own stability check, with a defined shelf life attached.
The candidate list, and where the comparisons disagree
The set that has actually been carried through peptide assembly rather than merely proposed includes Cyrene (dihydrolevoglucosenone), sulfolane, anisole, and dimethyl and diethyl carbonate [12]; ethyl acetate, 1,3-dioxolane and 2-methyltetrahydrofuran [4]; dimethyl sulfoxide as the polar partner [3, 10]; γ-valerolactone [11]; dipropyleneglycol dimethylether [5]; and NBP [1, 4]. Most are used in pairs rather than neat. In the carbonate work, mixtures of Cyrene, sulfolane or anisole with dimethyl or diethyl carbonate were assessed for their ability to swell differently functionalised polystyrene and polyethylene glycol resins and for amino acid and coupling reagent solubility, then taken through Aib-enkephalin as a model, the longer Aib-ACP, and the reduced form of octreotide [12].
DMM is a useful worked example of how resin-dependent these comparisons are. Assembled on Rink amide ChemMatrix, the Aib-enkephalin pentapeptide gave 94.88% HPLC purity in DMF and 95.60% in DMM; on Wang polystyrene the same peptide gave 89.20% in DMF and 81.76% in DMM. The Aib-ACP decapeptide, on Rink amide ChemMatrix, came out at 53.73% in DMF and 53.69% in DMM [5]. Read those three comparisons together: the solvent is interchangeable on one resin, worse on another, and irrelevant to the outcome on a sequence where something other than solvation is limiting. A single number quoted for “DMM versus DMF” means nothing without the resin and the sequence attached to it.
NBP shows why single head-to-head results should be read narrowly. One group reported that NBP gave a lower or equal level of racemisation for the amino acids most prone to it, and clearly lower aspartimide formation, than DMF — hence the claim that a green solvent can be synthetically better rather than merely tolerable [1]. A different group, screening Cyrene, ethyl acetate, 1,3-dioxolane, 2-methyltetrahydrofuran and NBP through microwave-assisted SPPS across five peptides including ACP(65–74), peptide 18A, thymosin α1 and the Jung–Redemann peptide, concluded that NBP performed on par with DMF in all tested syntheses, with NBP/ethyl acetate and NBP/2-methyltetrahydrofuran combinations also giving excellent results [4]. “Better” and “on par” are not contradictory — different heating regimes, resins, sequences and endpoints were interrogated — but neither is a consensus, and neither licenses a general claim that NBP outperforms DMF.
DMSO-containing mixtures keep recurring because DMSO carries the polarity budget while the partner supplies lower viscosity and the swelling. Screened on an automated synthesiser with controlled induction heating between 25 and 90 °C, binary mixtures containing dimethyl sulfoxide solubilised reagents efficiently while maintaining the desired swelling characteristics of common resins, and were assessed on ACP(65–74) and on longer, aggregating sequences [3].
Taking swelling out of the problem
An alternative reading is that the field has been optimising the wrong variable. Efforts to green SPPS have largely focused on identifying a single solvent capable of replacing DMF while relying exclusively on polystyrene-based resins, an approach described by its critics as both challenging and limiting [13]. Using a relatively rigid support changes the terms: when resin swelling is no longer a critical parameter, solvents that do not swell the resin become usable, and a rigid macroporous polyacrylate resin has been used to assemble small- to medium-sized peptides in ethyl acetate, TEP, 2-methyltetrahydrofuran, acetone and acetonitrile as single solvents [13]. Treat this as a promising reframing rather than a settled result — one laboratory, a new support, short to medium sequences — but it correctly identifies that “which solvent replaces DMF” is a question conditioned on the resin, not a property of the solvent alone.
Water
Water content is normally treated as a contaminant to exclude, but it can also be the medium. Fmoc/t-Bu SPPS has been run with standard Fmoc amino acids on TentaGel S in a 4:1 mixture of water with the green solvent PolarClean, where the resin/solvent combination played a dual role by improving both resin swelling and the solubility of starting materials [14]. In a model coupling, TCFH and 2,4,6-collidine gave full conversion at only 1.3 equivalents of amino acid, and Leu-enkephalin amide was obtained in 85% yield and 86% purity [14]. Waste was recycled by filtration through a mixed ion exchange resin and reused without affecting peptide quality [14]. The demonstrated scope is a pentapeptide, so this is a proof of principle rather than a general method — but it establishes that the water intolerance of Fmoc SPPS is a property of particular reagent sets, not an axiom of the chemistry.
PMI, the E factor, and what they miss
Both headline metrics are mass ratios. PMI considers all the materials used in a process and is calculated as the ratio between the total mass of materials and the mass of isolated product [5]. The E factor comes at the same accounting from the waste side, waste being defined as everything but the desired product [15]. The ACS Green Chemistry Institute Pharmaceutical Roundtable chose PMI as its key, high-level metric for evaluating and benchmarking progress towards more sustainable manufacturing, in preference to the E factor or atom economy [16].
What these numbers capture well is where the mass sits. The Roundtable analysis divided the peptide manufacturing process into synthesis, purification and isolation stages and determined the PMI of each [2], which is the only way to know whether a solvent substitution in the coupling step is even addressing the dominant term. They also capture recovery honestly. Recyclability is a selection criterion in its own right, and it is one on which a green candidate can win on process grounds rather than toxicological ones: DMM was recovered by distillation at 80% of all solvent used, and the PMI of that route fell from 3750.20 to 1370.68 through recycling [5].
What mass metrics do not capture is hazard. A kilogram of DMF and a kilogram of a benign ether contribute identically, so a substitution driven entirely by reprotoxicity can leave the metric unmoved — or worsen it, if the replacement needs more washes or a longer deprotection. Ranking solvents on safety, health and environment grounds requires a separate instrument: the CHEM21 guide built exactly that, proposing a set of Safety, Health and Environment criteria aligned with the Globally Harmonised System and European regulations, combined into an overall preliminary ranking of any solvent [17]. Nor does PMI say anything about crude purity. A route with a lower synthesis-stage PMI that halves the crude purity has moved mass into the purification stage, not removed it.
What is actually established
Reasonably firm: DMF's dual role can be reproduced by binary green mixtures selected on polarity and viscosity, and those mixtures did not adversely affect the purity profile relative to DMF on the sequences tested [8, 9]. Fmoc removal, not coupling, is the polarity-limited step, and changing the base is a legitimate way to widen the accessible solvent window [10]. Swelling depends more on linker and loading than on bead size, and does not blend linearly across a binary mixture [6, 7].
Not settled: which mixture is best, on which resin, at which scale. Most published comparisons are single-laboratory studies on different resins, with different coupling chemistry, different heating and small sequence panels — which is precisely the situation in which two groups can both be right about NBP and still disagree about the headline [1, 4]. The reframing toward rigid, low-swelling supports [13] and toward aqueous media [14] may make the neat-solvent question partly obsolete, but neither carries the breadth of sequence data that DMF has accumulated. If you are moving a protocol off DMF, the defensible procedure is narrow: measure swelling on your resin at your loading, measure Fmoc-removal kinetics in the candidate solvent with your actual base, confirm the base does not react with the solvent over the working life of the stock solution, and re-run your worst sequence rather than your easiest one before you believe any metric.