Electrospray Ionization (ESI)

ESI is the dominant ionization source for lc-ms-based bioanalysis, pharmacokinetics, and metabolomics: a chromatographic eluent flows directly into the electrospray needle, so every separated peak is ionized and mass-analyzed online, a workflow used for everything from routine drug-metabolite identification in pharmaceutical development to trace pesticide residue quantitation in food safety testing. In structural biology, "native ESI-MS" — spraying proteins from volatile, near-physiological buffers such as ammonium acetate — preserves non-covalent quaternary structure well enough to weigh entire intact antibody complexes, ribosomal subunits, or viral capsids, and to count how many ligand or drug molecules are bound per protein by the mass shift of the whole charge-state envelope. Intact-mass analysis of monoclonal antibody biotherapeutics (a $\sim$150 kDa glycoprotein) by ESI-MS is now a standard release-testing method in biopharmaceutical manufacturing, checking for glycoform heterogeneity, oxidation, and clipped variants directly from the multiply charged spectrum. Top-down proteomics — fragmenting an intact multiply charged protein ion inside the mass spectrometer to read out sequence and localize post-translational modifications without first digesting it into peptides — is built entirely on ESI's ability to deliver large, intact, highly charged precursor ions. Nanoelectrospray (nanoESI), operating at flow rates of tens of nanoliters per minute from pulled glass or metal-coated capillary tips, extends sensitivity down to single-cell and sub-picomole sample amounts, and is the ionization method underlying most modern shotgun proteomics pipelines. Downstream of every ESI charge-state envelope sits the deconvolution mathematics developed further in multiply-charged-ions-and-deconvolution, which is what actually converts the raw multiply charged spectrum into the single reported molecular weight.

:::quiz[A protein with M = 14,305.0 Da (average mass) shows an ESI peak at m/z = 1193.9 Th. What charge state is this, using $n = M/[(m/z) - m_p]$?]

• n = 10
• n = 12 ✓
• n = 14
• n = 16

:::faq[Do adduct ions like [M+Na]⁺ mean my sample is contaminated?] Not necessarily, and in trace amounts they are nearly unavoidable — sodium and potassium are ubiquitous in glassware, solvents, and even ambient dust, so [M+Na]⁺ and [M+K]⁺ peaks appear in almost every positive-ion ESI spectrum to some degree. A small adduct peak alongside a dominant [M+H]⁺ is normal and often useful, since its known mass offset (adding Na⁺, 22.99 Da, in place of H⁺, 1.007 Da, is a net shift of +21.98 Da) provides an internal cross-check on the assigned neutral mass. Large or dominant adduct peaks, however, do indicate that the sample or mobile phase carries excess non-volatile salt, and this is worth addressing (via desalting or a volatile-buffer LC method) because heavy salt adduction and ion suppression genuinely degrade signal-to-noise for the protonated species you usually want to quantify.

Ask the tutor: How exactly does the charge-state envelope from ESI get turned into a single deconvoluted molecular weight, and what happens when two different proteins' envelopes overlap?

Electrospray begins with an ordinary liquid — analyte dissolved in a polar solvent, typically water/methanol or water/acetonitrile with a small percentage of acid (formic or acetic) to promote protonation — pumped through a narrow metal capillary (typically 50–150 μm i.d.) at flow rates from tens of nL/min (nanospray) up to hundreds of μL/min (conventional ESI). A potential difference of 2–5 kV is applied between the capillary tip and a counter-electrode (the mass spectrometer's inlet), roughly 1 cm away, generating a field on the order of 10⁶ V/m at the liquid meniscus. That field does two things simultaneously. First, it polarizes and then electrolytically separates charge within the conductive liquid: in positive-ion mode, cations accumulate at the liquid surface facing the counter-electrode while electrons flow away through the capillary wall (an electrochemical oxidation occurs at the metal–liquid contact, making the ESI source formally a controlled-current electrolytic cell). Second, the accumulated surface charge feels an outward Coulombic force from the applied field, which competes against the liquid's surface tension trying to pull it back into a sphere. When the field is strong enough, the meniscus deforms from its resting hemispherical shape into a cone — the Taylor cone — with a half-angle very close to the theoretical value of 49.3°, derived by Sir Geoffrey Taylor in 1964 for a conducting fluid drop in an external field. At the cone's apex the field intensifies further (the tip radius shrinks), and above a threshold voltage the apex destabilizes into a thin jet, tens of micrometers or less in diameter, that breaks up (Rayleigh jet instability) into a stream of small, highly charged droplets — the primary droplets typically carry radii of order 1–3 μm and net charges close to (but somewhat below) their Rayleigh limit. These droplets are then swept by the electric field and any counter-flowing drying gas toward the mass spectrometer inlet, and it is during this flight that the crucial physics of desolvation and ion release occurs.

As a charged droplet travels through the warm, dry curtain gas region toward the inlet, solvent evaporates from its surface, shrinking its radius while its net charge stays essentially fixed (charge loss occurs only in discrete fission events, not by continuous leakage). Because surface charge repels itself while surface tension holds the droplet together, this shrinkage pushes the droplet toward instability. The Rayleigh limit is the charge $q_R$ at which the destabilizing Coulomb repulsion exactly equals the stabilizing surface-tension force for a given radius $r$, and once evaporation shrinks $r$ enough that the actual charge on the droplet exceeds $q_R(r)$, the droplet undergoes Coulombic fission: it emits a jet of much smaller, highly charged "offspring" droplets, typically carrying only ~2% of the parent's mass but a disproportionate ~15–20% of its charge. This asymmetric fission cascade repeats, generating a hierarchy of ever-smaller, ever more charge-dense droplets — and it is the last generation of droplets, nanometers in diameter, from which gas-phase ions finally emerge. Two mechanisms describe that final emergence. In the ion evaporation model (IEM), proposed by Iribarne and Thomson (1976), once a droplet becomes small enough (radius below roughly 10 nm) and its surface field strong enough, a solvated small ion can be field-desorbed directly from the droplet surface into the gas phase, essentially by tunneling over/through a reduced activation barrier at the highly curved, highly charged interface — this dominates for small, robust ions (alkali metal ions, small organic molecules, small peptides). In the charged residue model (CRM), proposed earlier by Dole (1968), the droplet instead keeps shrinking by evaporation and fissioning all the way down until it contains, on average, a single analyte molecule; as the last solvent molecules leave, the droplet's residual charge is deposited directly onto that molecule's surface (protonating available basic sites), producing the gas-phase ion. CRM is now understood to dominate for large, structured biomolecules such as folded proteins, because their size and rigidity make them poor candidates for direct field desorption but excellent templates for a shrinking-droplet endgame. Native mass spectrometry of folded proteins and non-covalent complexes relies specifically on CRM-type desolvation preserving quaternary structure.

The net phenomenology visible at the detector reflects both mechanisms and the underlying solution chemistry. In positive-ion mode, molecules with basic sites (amines, imidazoles) pick up one or more protons to form $[\mathrm{M}+n\mathrm{H}]^{n+}$; alkali metal or ammonium adducts, $[\mathrm{M}+\mathrm{Na}]^+$, $[\mathrm{M}+\mathrm{K}]^+$, $[\mathrm{M}+\mathrm{NH_4}]^+$, appear when those cations are present in the mobile phase or the sample matrix. In negative-ion mode, acidic analytes lose a proton to form $[\mathrm{M}-n\mathrm{H}]^{n-}$, or capture an anion such as chloride, acetate, or formate to give $[\mathrm{M}+\mathrm{Cl}]^-$, $[\mathrm{M}+\mathrm{CHOO}]^-$. For large biomolecules with many ionizable sites, ESI produces a charge-state envelope — a family of peaks $[\mathrm{M}+n\mathrm{H}]^{n+}$ for a range of $n$ — rather than a single peak, and it is exactly this envelope that later gets deconvoluted into a single neutral mass.

1. The Rayleigh limit for a charged droplet. A spherical liquid droplet of radius $r$ and surface tension $\gamma$ becomes electrostatically unstable to fission when its charge $q$ satisfies

where: $q_R$ = the Rayleigh limiting charge (C), $\varepsilon_0$ = vacuum permittivity ($8.854\times10^{-12}\ \mathrm{F\,m^{-1}}$), $\gamma$ = liquid surface tension (N/m), $r$ = droplet radius (m). Reading: the limiting charge scales as $r^{3/2}$, so as a droplet evaporates and shrinks, the charge it can stably hold falls faster than its actual (frozen) charge does — the droplet is driven toward instability purely by solvent loss, with no need to gain charge from anywhere.

2. The number of elementary charges at the Rayleigh limit.

where: $z_R$ = number of elementary charges (dimensionless), $e$ = elementary charge ($1.602\times10^{-19}$ C). Reading: for water ($\gamma \approx 0.072$ N/m) this gives a compact numerical rule of thumb, $z_R \approx 8\big(r/\text{nm}\big)^{3/2}$ for $r$ in nanometers — a 100 nm water droplet can stably carry roughly 8000 elementary charges before fissioning.

3. Mass-to-charge relation for multiply charged ESI ions.

where: $M$ = neutral analyte monoisotopic or average mass (Da), $n$ = number of charges (integer), $m_{\text{adduct}}$ = mass added per charge (for protonation, the mass of H⁺, 1.00728 Da), $m/z$ = the mass-to-charge ratio measured by the analyzer (Th, thomson). Reading: this single equation is why ESI can put a 66 kDa protein (bovine serum albumin) inside an analyzer with an m/z range of only a few thousand — sufficient protons ($n \sim 20$–$60$) divide the mass down into the accessible window, and reading off two adjacent charge states lets you solve for both $M$ and $n$ simultaneously (worked out below).

Example 1 (foundational): Rayleigh limit of a shrinking water droplet. An electrosprayed aqueous droplet has an initial radius of $r_0 = 1.50\ \mu\text{m} = 1.50\times10^{-6}$ m, carrying $q_0 = 1.00\times10^{-13}$ C of charge (fixed as it evaporates). Water's surface tension at the operating temperature is $\gamma = 0.0720\ \mathrm{N/m}$. Find the radius at which the droplet reaches its Rayleigh limit, and how many elementary charges that corresponds to. Set $q_0 = q_R(r) = 8\pi\sqrt{\varepsilon_0\gamma r^3}$ and solve for $r$:

Compute the denominator first:

Numerator: $q_0^2 = (1.00\times10^{-13})^2 = 1.00\times10^{-26}\ \mathrm{C^2}$.

Taking the cube root:

This is larger than the initial radius (1.50 μm), which means the droplet is already above its Rayleigh limit at $r_0$ — it must have shed charge already, or (more physically for this problem as posed) the given charge corresponds to a droplet that would only reach criticality by growing, which cannot happen by evaporation. This is a deliberately instructive check: re-reading the numbers, a droplet losing solvent at fixed charge moves from large $r$ toward small $r$, so instability is reached when $r$ falls to the critical value computed from $q_R(r) = q_0$ only if $q_0$ was sub-critical at $r_0$. Let's verify directly by computing $q_R$ at $r_0 = 1.50\ \mu\text{m}$:

Inside the square root: $(1.50\times10^{-6})^3 = 3.375\times10^{-18}\ \mathrm{m^3}$, so

Since $q_0 = 1.00\times10^{-13}\ \mathrm{C} > q_R(r_0) = 3.69\times10^{-14}\ \mathrm{C}$, the droplet as specified is already super-Rayleigh at 1.50 μm — meaning a real droplet carrying that much charge would have fissioned before reaching 1.50 μm, at some larger radius. Solving $q_0 = q_R(r)$ properly (as done above) gives that larger critical radius, $r_{\text{crit}} = 2.92\ \mu\text{m}$, confirming self-consistently that fission would already have occurred while the droplet was still larger than 1.50 μm. The number of elementary charges carried at that point is

The lesson: real ESI droplets never actually carry charge far above the Rayleigh limit for their instantaneous size — the physical picture is that a freshly formed primary droplet starts below $q_R$ and evaporation continuously shrinks $r$ (and hence $q_R(r)$) until it catches up to the fixed charge, triggering fission; the numbers above simply demonstrate how sensitively $q_R$ depends on $r^{3/2}$.

Example 2 (applied): resolving myoglobin's charge-state envelope. An ESI mass spectrum of horse heart myoglobin (average mass known from sequence to be $M = 16{,}951.5$ Da) shows a peak at $m/z = 1060.8$ Th. Which charge state does this correspond to, and what is the expected m/z of the adjacent (one-higher-charge) peak? Using $m/z = (M + n\,m_p)/n$ with $m_p = 1.00728$ Da, rearrange for $n$:

This rounds cleanly to $n = 16$, the $[\mathrm{M}+16\mathrm{H}]^{16+}$ ion — a very typical charge state for a compact folded protein of this size under native-like ESI conditions. Check self-consistency by computing $m/z$ directly at $n=16$:

This is close to (though not exactly, due to rounding of the quoted 1060.8 in the problem and real instrumental calibration tolerance) the observed value — the small residual (0.3 Th) is well within typical low-resolution quadrupole mass accuracy and does not affect charge-state assignment. Now find the $n=17$ peak expected nearby:

So the spectrum should show a companion peak at 998.2 Th (17+), 62.3 Th away from the 16+ peak — this progressively shrinking peak spacing at higher n is the classic "picket fence" appearance of a native protein charge-state envelope, and it is exactly this spacing that the deconvolution algorithms exploit.

Example 3 (exam/competition level): full charge-pair deconvolution from raw peak positions, no prior mass knowledge. An unknown protein gives two adjacent, unassigned ESI peaks at $m_1 = 1104.90$ Th and $m_2 = 1035.30$ Th (with $m_2$ corresponding to one more charge than $m_1$, i.e., $m_2 < m_1$). Determine the charge states and the neutral monoisotopic-equivalent mass, then verify against a third peak expected one charge further, and comment on required resolving power. Using the derived formula,

Compute the denominator: $m_1 - m_2 = 1104.90 - 1035.30 = 69.60$ Th. Compute the numerator: $m_2 - m_p = 1035.30 - 1.00728 = 1034.29$.

Rounding to the nearest physically meaningful integer, $n = 15$ for the lower-m/z peak ($m_2$), which makes the higher-m/z peak ($m_1$) the $n = 14$ state. Recompute $M$ from the $n=14$ equation for a consistency check:

Cross-check using the $n=15$ equation:

These two independent estimates disagree (15,454.5 Da vs. 15,514.4 Da, a 60 Da discrepancy) — a red flag that the rounding of $n = 14.86$ to an integer needs re-examination, since real data should give near-identical $M$ from both equations when $n$ is correctly assigned. Re-solving more carefully: because the raw (non-integer) $n=14.86$ sits closer to 15 than 14 is misleading here, since the correct interpretation is that $m_1$ and $m_2$ must differ by exactly one unit of charge for the derivation to apply; check instead by testing both plausible integer pairs directly against $M = n(m_1-m_p)=(n+1)(m_2-m_p)$. Testing $n=14$ (so $m_1$ is 14+, $m_2$ is 15+):

Testing $n=15$ (so $m_1$ is 15+, $m_2$ is 16+):

Testing $n=13$ (so $m_1$ is 13+, $m_2$ is 14+):

The smallest mismatch occurs at $n=15/16$, but none is exactly zero because these are synthetic peak values rounded to two decimals — in real data one instead uses a least-squares fit across many charge states simultaneously (the standard approach in charge-state deconvolution software), which averages out exactly this kind of rounding sensitivity. Taking the best (minimum-residual) integer assignment, $n=15$ for $m_1$, $n=16$ for $m_2$, and averaging the two mass estimates,

The pedagogical point: with only two adjacent peaks and finite decimal precision, small measurement error in m/z (here, effectively ±0.05–0.1 Th from rounding) propagates into a mass uncertainty of tens of daltons because the denominator $m_1-m_2$ is a small difference of two much larger numbers. This is precisely why real charge-state deconvolution (as used going into multiply-charged-ions-and-deconvolution) fits five, ten, or more charge states simultaneously by least squares — the redundancy suppresses the error amplification inherent in using just one adjacent pair, and typical native ESI-MS mass accuracy for a 16 kDa protein from a well-resolved envelope is better than ±1–2 Da.

John Bennett Fenn (1917–2010) spent much of his career working on molecular beams — supersonic jets of gas-phase molecules used to study collision dynamics — before turning, relatively late in his professional life, to the problem of getting large, fragile, non-volatile biomolecules into the gas phase for mass spectrometry. Working at Yale University in the 1980s, Fenn built on the earlier, largely unrecognized electrospray work of Malcolm Dole (who had proposed the charged residue mechanism and demonstrated electrospraying of polystyrene in 1968) and combined it with his own molecular-beam expertise to show, decisively, that intact proteins — including large ones like bovine albumin — could be sprayed from solution and detected as multiply charged ions with mass spectrometry. His key 1988–1989 papers demonstrated that this multiply charged ion series could be deconvoluted into accurate molecular weights for proteins previously inaccessible to MS. Fenn shared the 2002 Nobel Prize in Chemistry with Koichi Tanaka (for soft laser desorption, the precursor to MALDI) and Kurt Wüthrich (for NMR of biological macromolecules), a joint recognition of the moment mass spectrometry and structural biology of large molecules became routine. Fenn remained an outspoken, somewhat irreverent scientist well into his 80s and 90s, still publishing and lecturing on electrospray mechanism until shortly before his death at age 93.