Bound States and Zero-Energy Resonances

Scattering data live at $E>0$ ; bound states live at $E<0$ . They look like different subjects — one is measured with beams and counters, the other with spectroscopy — yet one analytic function knows both. This closing article of the SCAT core proves it: continue the partial-wave S-matrix $S_0(k)$ off the real axis and every bound state announces itself as a pole at $k=i\kappa$ on the positive imaginary axis. The derivation is the one Spherical Hankel Functions and Outgoing Waves explicitly relocated here ("the base's pole derivation is relocated there"), and it lands in three lines once the exterior wavefunction is written in travelling waves. The pole language then pays for itself. Poles on the negative imaginary axis are virtual states — near-misses with no normalizable wavefunction that nonetheless dominate low-energy scattering; complex pole pairs are the resonances of Resonances and the Breit-Wigner Formula; and a pole crossing $k=0$ is exactly the divergence of the scattering length seen on the universal curve of Low-Energy Scattering and the Scattering Length. We track that migration numerically for a square well — virtual pole climbing the axis, crossing threshold at the predicted depth, becoming the bound state — and we run the flagship arithmetic both ways: the deuteron's binding energy predicts $a_t=5.42$ fm, while the np singlet's $a_s=-23.7$ fm places its pole at $k=-0.040i$ fm $^{-1}$ — a virtual level about $66$ keV into the second sheet, the reason thermal neutrons scatter at $71$ barns off a state that does not exist. Half a century of physics — Levinson's theorem, halo nuclei, Efimov towers, Feshbach engineering — is bookkeeping on this plane.

Key takeaways

The s-wave S-matrix from matching is $S_0(k)=e^{-2ikR}\,\dfrac{\gamma+ik}{\gamma-ik}$ ( $\gamma$ = interior log-derivative): unimodular on the real axis, meromorphic off it.
Pole theorem (derived here, cashing the relocation from Spherical Hankel Functions and Outgoing Waves): $S_0$ has a pole at $k=i\kappa$ exactly when $K\cot KR=-\kappa$ — the bound-state matching condition of Particle in a Spherical Well and Infinite Spherical Box. Bound states are S-matrix poles.
The pole atlas: positive imaginary axis = bound ( $e^{-\kappa r}$ , normalizable) · negative imaginary axis = virtual ( $e^{+\kappa r}$ , not a state) · complex pairs in the lower half-plane = resonances ( $E_R-i\Gamma/2$ , owner Resonances and the Breit-Wigner Formula).
Near threshold the pole is universal: $k\cot\delta_0=-1/a$ puts it at $k=i/a$ , giving $E_b=\hbar^2/(2ma^2)$ — deuteron estimate $1.41$ MeV from $a_t$ alone (63% of truth), $2.22$ MeV once $r_t$ enters.
Tracked numerically: deepening a square well drives the virtual pole up the imaginary axis, through $k=0$ at $V_0R^2=\pi^2\hbar^2/8m$ (where $a\to\pm\infty$ — the zero-energy resonance), onto the bound axis.
Flagship: np singlet $(a_s,r_s)=(-23.7,\,2.75)$ fm $\Rightarrow$ pole at $\kappa_v=-0.040$ fm $^{-1}$ , a virtual state at $E_v\approx66$ keV on the second sheet — enormous scattering ( $4\pi a_s^2=71$ b), no binding.
Levinson bookkeeping: $\delta_0(0)-\delta_0(\infty)=n_b\pi$ — the staircase Phase Shifts exhibited, counted here by poles (measured: $0,\,0.99\pi,\,1.92\pi$ for wells holding $0,1,2$ states).
Key idea: bound states, virtual states, and resonances are all poles of the analytically continued S-matrix $S_\ell(k)$ ; their location in the complex- $k$ plane determines their physical character, and a pole approaching $k=0$ from either side makes the scattering length diverge.

The core insight

Bound states, virtual states, and resonances are all poles of the analytically continued S-matrix $S_\ell(k)$ ; their location in the complex- $k$ plane (positive imaginary, negative imaginary, or complex with negative imaginary part) determines their physical character, and a pole approaching $k=0$ from either side makes the scattering length diverge.

Where this lives today

Feshbach-resonance experiments are pole engineering, nothing less: a magnetic field slides a molecular level through an atomic threshold, dragging the associated pole along the imaginary axis exactly as our square-well table does with $V_0$ — and every landmark of the migration (halo dimers on one side, huge negative lengths on the other, unitarity in between) is read out in a single apparatus. Halo nuclei live in the shallow-bound corner: ¹¹Li's two loose neutrons, and the deuteron itself, are poles hugging $k=0$ , which is why their sizes and reactions obey universal formulas indifferent to nuclear detail. At exact unitarity ( $1/a=0$ , pole parked at the origin) three-body physics erupts into the Efimov tower — bound states in geometric progression with ratio $e^{2\pi/s_0}\approx515$ — observed in cold cesium and helium. And in hadron spectroscopy the same taxonomy adjudicates modern controversies: whether an enhancement like the $X(3872)$ or a pentaquark candidate is a bound state, a virtual state, or a resonance is, verbatim, the question "on which sheet does the pole sit?" — answered by fitting pole positions to scattering data, the trade this article teaches.

The physical picture

Take the exterior s-wave that every article in this chapter uses — incoming plus outgoing spherical waves — and ask an innocent question: what happens to the outgoing coefficient if I dial the energy negative? At $E<0$ the momentum turns imaginary, $k=i\kappa$ ; the "incoming" wave $e^{-ikr}$ becomes a growing exponential $e^{+\kappa r}$ and the "outgoing" wave becomes the decaying tail $e^{-\kappa r}$ that Spherical Hankel Functions and Outgoing Waves identified as the universal bound-state exterior. A bound state is a solution that is purely decaying — no growing piece — and the only way the exterior can shed its growing piece is for its coefficient ratio, which is exactly $S_0(k)$ , to blow up. A bound state is not "related to" a pole. It is a pole. That observation turns the complex $k$ -plane into a map of everything a potential can do:

Figure

The pole atlas: one complex k-plane holds bound states, virtual states, resonances, and scattering

Why does one plane hold two energy sheets? Because $E=\hbar^2k^2/2m$ is two-to-one: $k=+i\kappa$ and $k=-i\kappa$ map to the same negative energy. The $k$ -plane is the two-sheeted energy surface of Green Functions in Quantum Mechanics unfolded — its physical sheet is the upper half-plane, its second sheet the lower — which is why $k$ , not $E$ , is the honest variable for pole bookkeeping. A "virtual state at $66$ keV" and a "bound state at $66$ keV" would have the same energy; they differ in which $k$ -axis, hence which sheet, the pole occupies — and in everything physical that follows. The plane also animates. Deepen an attractive well and its nearest pole moves: up the negative imaginary axis, through the origin, onto the positive axis. The moment of crossing is the zero-energy resonance — the exterior wavefunction goes flat, the scattering length diverges, and the threshold cross section explodes. The divergences on the universal curve of Low-Energy Scattering and the Scattering Length are not a curiosity of the square-well formula; they are pole crossings, photographed from the real axis.

The S-matrix off the real axis

Every ingredient is already on the table. The exterior s-wave, in the travelling basis of Spherical Hankel Functions and Outgoing Waves, with the S-matrix packaging of The S-Matrix and Unitarity in Scattering:

u(r)\;\propto\; e^{-ikr}-S_0(k)\,e^{+ikr},\qquad r>R \qquad (1)

where: $u=rR_0$ is the reduced radial function, and $S_0=e^{2i\delta_0}$ on the real axis. Nothing stops us from evaluating $S_0$ at complex $k$ — the matching algebra below produces a ratio of entire functions, defined everywhere. Off the axis $|S_0|\neq1$ : unitarity is a real-axis promise only.

Derivation AThe pole theorem: a bound state is a pole of the S-matrix

We derive the closed form of $S_0(k)$ for any finite-range potential and prove that its poles on the positive imaginary axis are exactly the bound states — the derivation that Spherical Hankel Functions and Outgoing Waves relocated to this article by chapter ownership ("bound states as poles… RELOCATED there"), with provenance noted at both ends. The route: match the exterior (1) to the interior logarithmic derivative, solve for $S_0$ , then continue $k\to i\kappa$ and watch the boundary condition change its meaning. Watch one subtlety: the pole condition must reproduce, symbol for symbol, the $E<0$ matching condition derived independently in Particle in a Spherical Well and Infinite Spherical Box — that agreement is the theorem.

Step 1. Let $\gamma(k)$ be the logarithmic derivative $u'/u$ of the interior solution at $r=R$ (for the square well, $\gamma=K\cot KR$ with $K=\sqrt{K_0^2+k^2}$ , but the argument is general). Matching (1) at $r=R$ :

\frac{-ik\,e^{-ikR}-S_0\,ik\,e^{+ikR}}{e^{-ikR}-S_0\,e^{+ikR}}=\gamma \;\;\Longrightarrow\;\; S_0(k)=e^{-2ikR}\,\frac{\gamma+ik}{\gamma-ik} \qquad (2)

where: the numerator and denominator are analytic in $k$ ; $S_0$ is a meromorphic function whose poles sit where $\gamma(k)=ik$ . On the real axis $\gamma$ is real, numerator and denominator are complex conjugates, and $|S_0|=1$ — elastic unitarity recovered as a one-line check.

Step 2. Continue to the positive imaginary axis, $k=i\kappa$ with $\kappa>0$ . The pole condition $\gamma=ik$ becomes $\gamma=-\kappa$ ; for the square well, $K\cot KR=-\kappa$ — precisely the bound-state condition that Particle in a Spherical Well and Infinite Spherical Box derived by demanding $e^{-\kappa r}$ decay at $E<0$ . The two calculations never met until this line. Why the pole is forced physically. At $k=i\kappa$ equation (1) reads $u\propto e^{+\kappa r}-S_0\,e^{-\kappa r}$ . A bound state has no growing piece — and the only way a fixed combination can lose its $e^{+\kappa r}$ term is for the coefficient ratio $S_0$ to diverge, leaving pure decay. Pole = permission to drop the growing exponential = normalizability.

Step 3. Character of each axis. At the pole on the positive axis the surviving solution is $e^{-\kappa r}$ : square-integrable, a genuine eigenstate at $E=-\hbar^2\kappa^2/2m$ . A pole on the negative imaginary axis ( $k=-i\kappa_v$ , $\kappa_v>0$ ) leaves the surviving exterior $e^{+\kappa_v r}$ : growing, non-normalizable — not a state, yet still a pole of the same function, still commanding the low-energy amplitude nearby. That object is the virtual state.

Step 4. Zeros mirror poles: from (2), $S_0(-k)=1/S_0(k)$ ( $\gamma$ is even in $k$ for the square well and generally for real potentials), so each pole at $k_p$ pairs with a zero at $-k_p$ — the structure the Jost-function formalism makes systematic (named honestly: we state, and do not develop, that machinery). Verified: form (2) satisfies the matching identity symbolically; unimodularity on the real axis symbolic; the pole condition at $k=i\kappa$ reduces to $K\cot KR=-\kappa$ symbolically. Article verifier, 26 checks.

The theorem earns its keep when the pole sits near $k=0$ — details of $\gamma(k)$ stop mattering and the effective-range expansion takes over.

Go deeper · Derivation B · 2 minThe near-threshold pole is universal

Derivation BThe near-threshold pole is universal

We locate the pole using only the low-energy expansion $k\cot\delta_0=-1/a+\tfrac12r_0k^2$ of Low-Energy Scattering and the Scattering Length, deriving the universal binding law $E_b=\hbar^2/(2ma^2)$ and its effective-range correction — and, for $a<0$ , the virtual twin. The point of the derivation is what it does not use: no potential, no matching, no model. Any system with the same $(a,r_0)$ has the same near-threshold pole — nuclear, atomic, or engineered.

Step 1. Write $S_0$ in the amplitude variables of The S-Matrix and Unitarity in Scattering: $S_0=1+2ikf_0$ with $f_0=1/(k\cot\delta_0-ik)$ . Poles of $S_0$ are poles of $f_0$ :

k\cot\delta_0(k)=ik \qquad (3)

where: this equation continues analytically because $k\cot\delta_0$ is a function of $k^2$ (even in $k$ ) — the expansion variable of the effective-range line.

Step 2. Zero-range limit ( $r_0\to0$ ): $-1/a=ik$ , so $k_{\mathrm{pole}}=i/a$ . For $a>0$ the pole sits on the bound axis at $\kappa=1/a$ :

E_b=\frac{\hbar^2\kappa^2}{2m}=\frac{\hbar^2}{2ma^2} \qquad (4)

where: $m$ is the (reduced) mass. One measured length predicts a binding energy. For $a<0$ the same algebra puts the pole at $k=-i/|a|$ — the virtual axis. The sign of the scattering length is the sheet of the pole: that is the entire bound/virtual dictionary of the previous article, explained.

Step 3. First correction. Insert $k=i\kappa$ into the full line: $-1/a-\tfrac12r_0\kappa^2=-\kappa$ , a quadratic with roots

\kappa=\frac{1\pm\sqrt{1-2r_0/a}}{r_0} \qquad (5)

where: the root with the minus sign is the physical near-threshold pole (it reproduces $\kappa\to1/a$ as $r_0\to0$ ); the plus root sits at $\kappa\sim1/r_0$ , outside the expansion's reach — a spurious solution to be rejected on validity grounds, not kept. Why the honest rejection matters. The effective-range line is a two-term truncation valid for $|k|\lesssim1/r_0$ -scale. Its second root lies exactly at the edge of that validity, where the truncation itself is meaningless — quoting it as a physical prediction would be manufacturing a state out of an error term.

Step 4. The residue carries physics too: near the pole $f_0\approx-\dfrac{1}{a}\dfrac{1}{k^2+\kappa^2}$ -type, and the residue of $S_0$ at $k=i\kappa$ fixes the asymptotic normalization of the bound-state tail — the quantity nuclear physicists tabulate as the ANC. We state this and move on; the flagship numbers below are the real payoff. Verified: pole of the zero-range $S_0$ at $k=i/a$ symbolic; quadratic roots and branch selection numeric for the np channels; the ERE pole reproduces the exact square-well pole to 0.2% at $V_0=30$ MeV. Article verifier.

Watching the pole move

Theory says: deepen the well, the pole climbs. The verifier does it. For an np-flavored square well of radius $R=2$ fm, the pole of (2) is the root of $K\cot KR=-\kappa$ , tracked as $V_0$ rises through the binding threshold $V_0^*=\pi^2\hbar^2/8mR^2=25.6$ MeV:

$V_0$ (MeV)	pole $\kappa$ (fm $^{-1}$ )	character
15	$-0.308$	deep virtual
20	$-0.147$	virtual, approaching
24	$-0.039$	shallow virtual — huge $\sigma(0)$ , $a<0$ large
25.6	$0.000$	zero-energy resonance: $a=\pm\infty$
27	$+0.034$	shallow bound — halo
30	$+0.100$	bound
35	$+0.202$	bound, deuteron-like

At $V_0=25.4$ MeV the scattering length is $-227$ fm; at $25.8$ MeV it is $+191$ fm. The divergence of $a(V_0)$ and the pole's crossing of $k=0$ are the same event seen by two instruments — this is the punchline the previous article promised.

Figure

Punchline — the measured pole migration: virtual to zero-energy resonance to bound, with the scattering length exploding at the crossing

Levinson's theorem is this picture's census: each pole that has crossed onto the bound axis leaves a permanent $\pi$ in the zero-energy phase, $\delta_0(0)-\delta_0(\infty)=n_b\pi$ (Levinson 1949, stated with its source; the proof belongs to Jost-function theory, named and not performed). The staircase that Phase Shifts exhibited by deepening the well is the real-axis shadow of pole crossings — our verifier measures $-0.003\pi,\ 0.990\pi,\ 1.920\pi$ for wells holding $0,1,2$ states, the deficits from exact integers being finite- $k_{\max}$ honesty, not physics.

The flagship: one channel bound, one channel virtual

The neutron–proton system hands us both pole types in a single experiment, and the T-matrix face of the same statement was already glimpsed: the separable model of Lippmann-Schwinger Equation watched its denominator $1-\lambda g_0$ "head for zero at threshold — a bound state being born." Poles of $\hat T$ and poles of $S$ are one phenomenon; here they get measured coordinates.

Triplet channel. The deuteron binds at $E_B=2.2246$ MeV, so $\kappa=\sqrt{2\mu E_B}/\hbar c=0.2316$ fm $^{-1}$ — the tail Spherical Hankel Functions and Outgoing Waves computed. Run the pole condition (5) backwards with $r_t=1.76$ fm: predicted $a_t=1/(\kappa-\tfrac12r_t\kappa^2)=5.42$ fm, matching the measured $5.424$ fm. Forward, from $a_t$ alone, the universal estimate (4) gives $E_b=1.41$ MeV — 63% of the truth, the honest price of zero-range universality on a state whose $r_0/a\approx0.32$ .

Singlet channel. $(a_s,r_s)=(-23.75,\,2.75)$ fm. The physical root of (5) is $\kappa_v=-0.0399$ fm $^{-1}$ : negative — the pole sits on the virtual axis, $|E_v|=\hbar^2\kappa_v^2/2\mu\approx66$ keV into the second sheet ( $74$ keV in the $a$ -only estimate; the spread between the two numbers is the honest uncertainty of truncating the expansion). No bound state, no spectral line, no partner to the deuteron — yet $\sigma_s(0)=4\pi a_s^2=71$ b, dwarfing anything geometry allows. Almost-binding is a pole distance, and $0.04$ fm $^{-1}$ is very close indeed.

Limiting cases

case	result	meaning
pole at $k=i\kappa$ , $\kappa>0$	bound state, $E=-\hbar^2\kappa^2/2m$ , $u\propto e^{-\kappa r}$	normalizable; physical energy sheet
pole at $k=-i\kappa_v$	virtual state: same $E<0$ , second sheet	$e^{+\kappa_v r}$ exterior — not a state; commands nearby scattering
pole pair $\pm k_R-ik_I$	resonance $E_R-i\Gamma/2$	quasi-bound; line shape owned by Resonances and the Breit-Wigner Formula
pole $\to k=0$	$a\to\pm\infty$ , $\sigma(0)\to\infty$	zero-energy resonance; the universal curve's divergence
$\kappa\ll1/r_0$ (shallow)	$E_b=\hbar^2/2ma^2$ , corrections in $r_0/a$	universality: potential details forgotten
$\kappa\sim1/r_0$	expansion (5) breaks; plus-root spurious	pole leaves the universal window — model required
$k\to0$ exactly at threshold	$\delta_0(0)$ jumps by $\pi$ (Levinson census)	each crossing adds one bound state to the books

Common trapA virtual state is not a faint bound state — and its energy is not a level

Three precision failures haunt this subject. First, the virtual state has no wavefunction: $e^{+\kappa_v r}$ grows, nothing is normalizable, nothing sits in the spectrum — "state" is a courtesy title for a pole. Second, its "energy" $-\hbar^2\kappa_v^2/2m$ lies on the second sheet: no spectroscopy will find a line at $66$ keV in the np system; the number locates a pole, not a level (the two-sheet grammar is Green Functions in Quantum Mechanics's). Third, the popular shorthand "the singlet deuteron is slightly unbound" must be cashed carefully: the pole is real and its effects (71 b!) are real, but promoting it to an eigenstate — or using $m$ instead of the reduced mass $\mu$ in $E=\hbar^2\kappa^2/2m$ — produces numbers off by factors of 2 and claims off by a category.

Worked examples

Example 1 — the deuteron, both directions. Given: $E_B=2.2246$ MeV, $a_t=5.424$ fm, $r_t=1.759$ fm, $\mu c^2=469.46$ MeV, $\hbar c=197.33$ MeV·fm.

Step 1. Binding energy → pole: $\kappa=\sqrt{2\mu E_B}/\hbar c=\sqrt{2\times469.46\times2.2246}/197.33=0.2316$ fm $^{-1}$ . Tail $1/\kappa=4.32$ fm.

Step 2. Pole → scattering length, with the effective-range bridge: $a_t^{\mathrm{pred}}=\big(\kappa-\tfrac12r_t\kappa^2\big)^{-1}=(0.2316-0.0472)^{-1}=5.42$ fm. Measured: $5.424$ fm — sub-percent agreement. Answer: the bound-state pole predicts $a_t=5.42$ fm.

Step 3. Scattering length → binding energy, zero-range: $E_b=\hbar^2/(2\mu a_t^2)=(197.33)^2/(2\times469.46\times5.424^2)=1.41$ MeV — $63\%$ of the measured $2.2246$ . With the $r_t$ root of (5): $\kappa=0.2318$ fm $^{-1}$ , $E_b=2.22$ MeV — $0.1\%$ . Answer: $1.41$ MeV (universal), $2.22$ MeV (one correction). The lesson is quantitative: universality's accuracy is governed by $r_0\kappa\approx0.41$ .

Example 2 — locating a state that does not exist. Given the np singlet $(a_s,r_s)=(-23.749,\,2.75)$ fm, find the pole and its observable footprint.

Step 1. Roots of (5): $\sqrt{1-2r_s/a_s}=\sqrt{1.2316}=1.1098$ , so $\kappa=(1-1.1098)/2.75=-0.0399$ fm $^{-1}$ (physical root) or $+0.767$ fm $^{-1}$ (rejected: $|\kappa|\sim1/r_s$ , outside the expansion's validity).

Step 2. Negative $\kappa$ ⇒ pole at $k=-0.0399i$ fm $^{-1}$ : virtual axis. Second-sheet energy $|E_v|=(\hbar c)^2\kappa^2/2\mu c^2=(197.33\times0.0399)^2/(2\times469.46)=66$ keV ( $74$ keV from $a_s$ alone — quote the spread, it is the method's error bar).

Step 3. Footprint on the real axis: $\sigma_s(0)=4\pi a_s^2=7087$ fm $^2=70.9$ b, against a geometric $4\pi R^2\approx0.5$ b for a 2-fm force. Answer: pole at $\kappa_v=-0.040$ fm $^{-1}$ ( $\sim66$ keV, second sheet); a 140-fold cross-section amplification from a state with no wavefunction. Historically this is how it was settled: scattering magnitude alone cannot fix the sign of $a_s$ , but neutron scattering off ortho- versus para-hydrogen — where the two protons' amplitudes interfere with fixed relative spin — reads the sign directly (Schwinger and Teller's 1937 calculation; the postwar measurements confirmed $a_s<0$ : virtual, not bound).

Origins · 1 minVirtual State and how this was discovered

Virtual State

1935–1947 · Rome, Princeton, Washington

The virtual state was forced on physics by a number that refused to fit: thermal neutrons scattered off protons at tens of barns, far more than the newly measured deuteron could explain. Eugene Wigner read the excess as spin dependence — the singlet channel must hide a level of its own near zero energy — but nothing in spectroscopy showed it, bound or otherwise. The 1937 Schwinger–Teller calculation turned molecular hydrogen into an interferometer: in para-H₂ the two proton amplitudes add with opposite spin weights than in ortho-H₂, so the relative sign of triplet and singlet scattering lengths — invisible in any cross section off free protons — prints directly onto the ortho/para ratio. When the measurements matured in the 1940s the verdict was unambiguous: $a_s<0$ . The singlet "level" sits on the wrong axis — a pole with no state attached, christened virtual. Its modern descendants run from the di-neutron (virtual, by a whisker) to engineered ultracold poles parked at will on either side of threshold.

History

1935

— Wigner attributes the anomalously large thermal np cross section to spin dependence: a near-threshold singlet level, character unknown

1936

— Bohr's compound nucleus and Breit–Wigner's formula put resonance poles to work at positive energy

1937

— Schwinger and Teller compute neutron scattering off ortho- and para-hydrogen: molecular interference reads the sign of the singlet length

1947

— postwar ortho/para and coherent-scattering measurements settle it: the singlet np state is virtual, roughly 70 keV onto the second sheet

1949

— Levinson ties the zero-energy phase to the bound-state count, making pole bookkeeping a theorem

1970

— Efimov predicts the geometric tower of three-body states at diverging scattering length, ratio 515 per rung

Going further

Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 6 — S-matrix poles and low-energy scattering.
Taylor, Scattering Theory, Ch. 12–13; Newton, Scattering Theory of Waves and Particles, 2nd ed., Ch. 12 — Jost functions, analyticity, Levinson's theorem done properly.
Landau & Lifshitz, Quantum Mechanics, 3rd ed., §133 — virtual states and the np system.
Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. 2, Ch. VIII — square-well thresholds.
Braaten & Hammer, Phys. Rep. 428, 259 (2006) — universality at large scattering length and Efimov physics.

Provenance. Equations cross-checked against Sakurai & Napolitano Modern QM 2e Ch.6, Taylor Scattering Theory Ch.12–13, Newton Scattering Theory Ch.12, and Landau-Lifshitz QM §133. Verifier should confirm: (1) the S-matrix Jost-function form $S_\ell=f_\ell(-k)/f_\ell(k)$ and the bound-state pole at $k=i\kappa$ ; (2) the universal amplitude pole at $k=i/a$ and binding energy $E_b=-\hbar^2/2ma^2$ ; (3) the square-well scattering length $a=R[1-\tan(k_0R)/(k_0R)]$ and the first-bound-state threshold $k_0R=\pi/2$ ; (4) the unitarity-saturated cross section $\sigma_0=4\pi/k^2$ at $\delta_0=\pi/2$ ; (5) Levinson's theorem $\delta_\ell(0)-\delta_\ell(\infty)=n_\ell\pi$ . Numerical values (deuteron $E_b\approx2.22$ MeV, $a_t\approx5.4$ fm, $a_s\approx-23.7$ fm; Efimov ratio $\approx515.0$ with $s_0\approx1.00624$ ) are standard textbook/measured values. — Items (1)–(3) and (5) are executed in this article's verification script (26 checks: pole condition both routes, $k=i/a$ , thresholds, migration table, Levinson staircase, Efimov ratio); item (4) is The S-Matrix and Unitarity in Scattering's ceiling, recalled not re-derived; the Jost-function formalism itself is named with its owners (Taylor, Newton) and not developed here.

Builds on: The S-Matrix and Unitarity in Scattering · Phase Shifts · Partial-Wave Expansion · Asymptotic Form of the Scattering Wave Function Leads to: Optical Theorem · Differential and Total Cross Section · Scattering Amplitude

Is a virtual state just a resonance with zero width?

No — different geography. A resonance is a pair of poles off the imaginary axis at $\pm k_R-ik_I$ , giving a peak at positive energy $E_R$ with width $\Gamma$ ; you can sit a beam on it. A virtual state is a single pole on the negative imaginary axis: negative second-sheet energy, no peak at any positive energy — only a smooth, enormous enhancement of threshold scattering. In s-waves, shallow attractive systems generically produce virtual states rather than narrow resonances precisely because there is no centrifugal barrier to trap anything; the barrier-made resonances belong to Resonances and the Breit-Wigner Formula.

If the virtual state has no wavefunction, in what sense does it exist?

As an analytic feature of the same function that scattering measures. $S_0(k)$ on the real axis is experimental data; a pole at $k=-i\kappa_v$ close to the origin forces $|f_0|^2\approx1/(k^2+\kappa_v^2)$ on the real axis nearby — a Lorentzian-in- $k^2$ enhancement that detectors see directly (71 b in the np singlet). "Exists" here means: a parameter of the data with a universal, model-independent meaning. What it is not is an eigenstate — nothing decays into it, nothing populates it, no line appears at its energy.

Why does the sign of the scattering length decide bound versus virtual?

Because the zero-range pole condition is $-1/a=ik$ , i.e. $k=i/a$ : positive $a$ puts the pole on the positive (bound) axis, negative $a$ on the virtual axis. Geometrically, the zero-energy exterior line $u\propto(r-a)$ with $a>0$ can be smoothly bent into a decaying exponential (its intercept is outside the origin), while the $a<0$ line points the wrong way — the "state" it extrapolates to grows with $r$ . Same dictionary, wavefunction edition.

Can virtual states and these pole migrations happen for higher partial waves too?

The plane is the same but the traffic rules differ. For $\ell\ge1$ the centrifugal barrier makes the threshold behavior $\tan\delta_\ell\sim k^{2\ell+1}$ , and a pole approaching $k=0$ generically arrives as a narrow resonance (a pair sliding in from the lower half-plane) rather than sliding along the imaginary axis: the barrier traps the nearly-bound state, giving it a lifetime instead of virtuality. The s-wave is special exactly because nothing confines it at threshold — which is why "virtual state" stories (np singlet, di-neutron, cold-atom thresholds) are s-wave stories.

Do I use the nucleon mass or the reduced mass in E = ħ²κ²/2m for np poles?

Reduced mass, always — the pole lives in the relative coordinate. For np, $\mu c^2=469.46$ MeV, and using $m_pc^2\approx938$ would halve every $1/\kappa$ and double every energy: the deuteron's $\kappa$ would come out $0.33$ instead of $0.2316$ fm $^{-1}$ . The article's verifier computes $\mu$ from CODATA masses before anything else — copy that discipline.

What exactly is the second sheet, and why does the virtual pole live there?

$E=\hbar^2k^2/2m$ maps two $k$ -values to each $E$ , so functions of energy inherit a two-sheeted surface with a cut along $E>0$ — the resolvent grammar of Green Functions in Quantum Mechanics and Retarded and Advanced Green Functions. The upper-half $k$ -plane unfolds to the physical sheet (where hermiticity forbids complex poles), the lower-half to the second sheet (where it doesn't). The virtual pole at $k=-i\kappa_v$ has $\mathrm{Im}\,k<0$ : second sheet by definition. Same real energy as a bound state, opposite sheet — and only the physical sheet's poles are eigenvalues.

How do experimentalists actually decide bound vs virtual vs resonance for a new bump?

Fit the amplitude's analytic structure to data and locate the pole. In practice: measure $\sigma(E)$ (and phases where possible) near threshold, fit an effective-range or K-matrix parametrization, continue analytically, and report the pole position with uncertainties. The np singlet was settled this way with 1940s technology (sign of $a_s$ via ortho/para-H₂ interference); the modern X(3872) and pentaquark debates are the same protocol with amplitudes from LHCb — the answer is stated, verbatim, as "the pole lies on sheet II at…". Pole coordinates are the lingua franca because they are parametrization-independent; bump shapes are not.