Schmidt Number: Uses in Quantum & Fluid Mechanics

• The Schmidt Number is a measure that indicates entanglement (in quantum theory) and the ratio of kinematic viscosity to molecular diffusivity (in fluid mechanics). It determines the entanglement dimensionality in quantum mechanics as the minimum Schmidt rank of pure-state decompositions and can be used in high-gain parametric down-conversion to evaluate mode populations.
• The Schmidt number has applications in quantifying the minimum Schmidt rank required in pure-state ensembles for achieving mixed quantum states and defining entanglement-preserving activities in quantum communication channels, and assessing the scalar diffusivity in fluid mixtures.
• Calculating Schmidt numbers and interpreting results can play a critical role in experiments, especially in assessing quantum entanglement for multipartite systems with generalized ranges, and in forcing turbulent mixing in urban canopy models.

The Schmidt number is a dimensionless quantity with distinct meanings in quantum information, quantum optics, and fluid mechanics. In bipartite quantum theory, it quantifies the minimum Schmidt rank required in any pure-state ensemble realizing a mixed state, thereby measuring entanglement dimensionality. For quantum channels, it is defined through the Schmidt number of the Choi state. In high-gain parametric down-conversion, an effective Schmidt number is an inverse participation ratio of mode populations. In scalar transport, the molecular Schmidt number is the ratio of kinematic viscosity to molecular diffusivity, while the turbulent Schmidt number is the ratio of turbulent momentum diffusivity to turbulent mass diffusivity.

1. Quantum definition and entanglement hierarchy

For a bipartite pure state,

$$|\psi\rangle_{AB}=\sum_{j=1}^{r}\lambda_j|a_j\rangle_A|b_j\rangle_B,$$

the Schmidt rank is the number $r$ of nonzero Schmidt coefficients. Equivalently, in finite dimensions it is the rank of either reduced density operator. Schmidt rank $1$ characterizes product states, while larger Schmidt rank indicates entanglement involving more local degrees of freedom (Guo et al., 2012).

For a mixed state, the Schmidt number is the minimum, over all pure-state decompositions, of the largest Schmidt rank appearing in the decomposition: $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$ Thus, $\operatorname{SN}(\rho)\leq k$ if and only if $\rho$ admits a decomposition into pure states of Schmidt rank at most $k$. The associated convex sets are

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$

with

$$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$

The value $k=1$ is equivalent to separability. A state with Schmidt number $r$0 requires at least $r$1-dimensional entanglement in every pure-state ensemble realizing it. Schmidt number is therefore not the largest Schmidt rank appearing in an arbitrary decomposition; it is the smallest possible upper rank after optimization over all decompositions (Tavakoli et al., 2024).

For pure states, Schmidt number and Schmidt rank coincide. Schmidt number should also be distinguished from the numerical Schmidt coefficients: the former is an integer-valued entanglement-dimensionality measure, whereas the latter describe the distribution of amplitudes among the Schmidt modes. A state can have maximal Schmidt rank without being maximally entangled if its Schmidt coefficients are unequal (Guo et al., 2012).

The Schmidt-number hierarchy is monotone under local operations and classical communication: $r$2 It is invariant under invertible local operators. The hierarchy is generally neither convex nor concave as a numerical function of $r$3, although each set $r$4 is convex (Chen et al., 2016).

2. Channels, Choi states, and preservation

For a quantum channel $r$5 acting on a $r$6-level system, the Choi–Jamiołkowski state is

$r$7

where

$r$8

The Schmidt number of the channel is the Schmidt number of $r$9. It describes the entanglement remaining when one half of a maximally entangled state is transmitted through the channel. This process property differs from the Schmidt number of an arbitrary output state, which is a property of that particular bipartite state (Namiki et al., 2012).

Channel Schmidt number $1$0 is equivalent to a separable Choi state. Such channels are exactly the entanglement-breaking, or measure-and-prepare, channels. Their action has the form

$1$1

so the channel measures the input, records a classical outcome, and prepares an output state depending on that outcome. It cannot transmit entanglement.

More generally, let $1$2 be the set of channels with Schmidt number at most $1$3. Given input states $1$4, target states $1$5, and prior probabilities $1$6, define

$1$7

The channel benchmark is

$1$8

If an experiment obtains

$1$9

then the process has channel Schmidt number at least $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$0.

For a $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$1-level identity memory tested with two mutually unbiased bases, the combined fidelity is

$$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$2

where $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$3 tests storage of $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$4 orthogonal states and $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$5 tests preservation of coherent superpositions. The central benchmark is

$$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$6

Therefore,

$$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$7

The ordinary quantum benchmark is the $$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$8 case,

$$\operatorname{SN}(\rho) = \min_{\rho=\sum_i p_i|\psi_i\rangle\langle\psi_i|} \left\{ \max_i \operatorname{SR}(|\psi_i\rangle) \right\}.$$9

which equals $\operatorname{SN}(\rho)\leq k$0 for a qubit. The test uses $\operatorname{SN}(\rho)\leq k$1 input states and $\operatorname{SN}(\rho)\leq k$2 input-output measurement settings, compared with the order-$\operatorname{SN}(\rho)\leq k$3 settings required for complete process tomography (Namiki et al., 2012).

A related preservation theorem concerns local channels acting on bipartite systems. If a local channel preserves every pure state of one fixed Schmidt number $\operatorname{SN}(\rho)\leq k$4, preserving both purity and the value $\operatorname{SN}(\rho)\leq k$5, then each local component must be an isometry. In equal finite dimensions, the isometries are unitaries: $\operatorname{SN}(\rho)\leq k$6 The $\operatorname{SN}(\rho)\leq k$7 case has an additional possibility: a channel that maps every input to a fixed pure state can produce pure separable outputs, but it cannot preserve nontrivial entanglement (Guo et al., 2012).

3. Witnesses, positive maps, and certification

A Hermitian operator $\operatorname{SN}(\rho)\leq k$8 is a Schmidt-number witness of class $\operatorname{SN}(\rho)\leq k$9 if

$\rho$0

while

$\rho$1

for at least one state with Schmidt number greater than $\rho$2. A negative expectation value therefore certifies

$\rho$3

The dual formulation uses $\rho$4-positive maps. A map $\rho$5 is $\rho$6-positive if

$\rho$7

is positive. The Terhal–Horodecki characterization is

$\rho$8

The Choi correspondence identifies $\rho$9-positive maps with $k$0-block-positive operators, and $k$1-superpositive maps with operators of Schmidt number at most $k$2 (Park et al., 2023).

The maximally entangled overlap gives a basic witness criterion. For an $k$3 state with Schmidt number at most $k$4,

$k$5

Consequently, overlap larger than $k$6 proves Schmidt number greater than $k$7 (Chen et al., 2016).

Recent measurement-based criteria replace direct state reconstruction by trace norms of correlation matrices. For SIC-POVMs, if $k$8 is the joint-probability matrix and

$k$9

then

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$0

Violation certifies $$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$1. For complete mutually unbiased bases, the corresponding condition is

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$2

These criteria are sufficient rather than necessary: satisfying the inequality does not establish that the state has Schmidt number at most $$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$3 (Tavakoli et al., 2024).

General symmetric informationally complete measurements extend the result to arbitrary local dimensions and tunable measurement parameters. For GSIC correlation matrix $$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$4, define

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$5

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$6

and

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$7

Then

$$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$8

The construction reduces to the SIC criterion at the rank-one endpoint (Wang et al., 2024).

An MDI formulation removes the need to trust the measurement devices. In a semi-quantum game, trusted quantum input states are sent to untrusted measurement devices. For every state satisfying $$\mathcal S_k=\{\rho:\operatorname{SN}(\rho)\leq k\},$$9, there exists a game whose payoff is nonnegative for all states in $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$0 but negative for the target state. Thus every finite-dimensional bipartite state can, in principle, have its Schmidt number certified in an MDI manner. Fully device-independent Bell tests do not have this property: Bell-local states can be entangled, and states with Schmidt number $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$1 can have projective-measurement correlations reproducible by states with Schmidt number at most $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$2 (Mukherjee et al., 18 Feb 2025).

4. PPT entanglement, projections, and multipartite extensions

PPT entanglement provides an important distinction between Schmidt number and distillability. A state is PPT when

$$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$3

PPT entangled states are bound entangled and cannot yield pure-state entanglement through standard distillation protocols. Nevertheless, PPT does not impose a finite upper bound on Schmidt number. For every positive integer $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$4, there exists a PPT-entangled state with Schmidt number $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$5 (Chen et al., 2016).

Under a local projection

$$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$6

where $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$7 removes a $$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$8-dimensional local subspace, the Schmidt number satisfies

$$\mathcal S_1\subseteq\mathcal S_2\subseteq\cdots \subseteq\mathcal S_{\min(d_A,d_B)}.$$9

Thus a projection can reduce Schmidt number by no more than the dimension removed. If $k=1$0, then

$k=1$1

The generalized range criterion states that if $k=1$2, then the range of $k=1$3 must be spanned by vectors of Schmidt rank at most $k=1$4. Therefore, if the range contains a vector orthogonal to every rank-$k=1$5-or-lower vector in the range, then $k=1$6. This criterion is particularly effective for sparse grid states, where Schmidt-rank constraints reduce to vanishing minors of coefficient matrices (Krebs et al., 2024).

Explicit constructions include a Schmidt-number-three PPT state in $k=1$7, and families with

$k=1$8

for odd $k=1$9-dimensional systems, with the other local dimension

$r$00

These constructions show that undistillability and entanglement dimensionality describe different properties.

For multipartite systems, no universal Schmidt decomposition generally exists. One generalization recursively combines local ranks and Schmidt numbers of complementary reductions. For a genuinely tripartite state,

$r$01

where $r$02 is the rank of the one-party reduction and $r$03 is the bipartite Schmidt number of the complementary reduction. The corresponding mixed-state quantity is defined by minimizing the maximum generalized pure-state rank over all decompositions. The construction is an entanglement monotone under LOCC and is invariant under invertible SLOCC (Guo et al., 2013).

A distinct multipartite quantity is the joint Schmidt number, the tuple of Schmidt numbers across all one-versus-rest bipartitions: $r$04 If $r$05 is the tensor rank, then

$r$06

This separates global tensor rank from the entanglement structure visible across individual bipartitions (Chen et al., 2016).

5. Optical mode dimensionality

In parametric down-conversion, the Schmidt number also quantifies the effective number of correlated spatial or spectral modes. In the low-gain regime, a biphoton state has the form

$r$07

with effective Schmidt number

$r$08

This is the inverse participation ratio of the Schmidt-weight distribution. Equal occupation of $r$09 modes gives $r$10.

In high-gain bright squeezed vacuum, the effective mode weights depend on the parametric gain $r$11: $r$12 and

$r$13

Unlike $r$14, the high-gain Schmidt number $r$15 is gain-dependent. Increasing gain preferentially amplifies the most strongly coupled modes, so the effective mode number generally decreases (Dyakonov et al., 2014).

For a multimode thermal field, the single-beam second-order correlation satisfies

$r$16

hence

$r$17

If signal and idler contain the same effective number of modes, measuring one beam is sufficient to infer the bipartite effective mode number. The measurement uses a Hanbury Brown–Twiss interferometer and does not require simultaneous signal-idler coincidence detection.

In the reported high-gain experiment,

$r$18

with a theoretical spatial prediction

$r$19

The full-beam result remained approximately constant from the near field to the far field because free-space propagation is unitary and the full transverse spectrum was collected. Finite apertures instead measure a selected spatial subsystem and can produce a propagation-dependent effective mode number.

The optical effective Schmidt number should not automatically be identified with the low-gain entanglement dimensionality. It measures the gain-dependent population of modes in the bright squeezed vacuum and relies on equal effective mode numbers in the two beams, suitable thermal statistics, and complete or consistently defined collection.

6. Schmidt numbers in scalar and turbulent transport

In fluid mechanics, the molecular Schmidt number is

$r$20

where $r$21 is kinematic viscosity and $r$22 is molecular scalar diffusivity. It compares momentum diffusion with molecular scalar diffusion.

For $r$23, scalar diffusion is weak relative to momentum diffusion. Velocity fluctuations are smoothed at the Kolmogorov scale $r$24, while scalar fluctuations persist to the smaller Batchelor scale

$r$25

For $r$26, scalar diffusion is strong and the relevant scalar cutoff is the Corrsin–Obukhov scale

$r$27

At $r$28, scalar and velocity dissipative ranges overlap.

In forced compressible turbulence, the scalar spectrum follows a $r$29 law in the inertial-convective range. At high Schmidt number, a viscous-convective range with

$r$30

appears. At low Schmidt number, an inertial-diffusive range with

$r$31

is identified. High $r$32 produces roll-up, thin streamers, and fine-scale scalar structure, whereas low $r$33 erases small-scale structure and leaves broad, cloudlike regions (Ni, 2015).

The Schmidt number used in turbulence modeling is distinct from the molecular Schmidt number. The turbulent Schmidt number is

$r$34

where $r$35 is turbulent eddy diffusivity of momentum and $r$36 is turbulent eddy diffusivity of mass or a passive scalar. In a gradient-transport closure,

$r$37

Thus $r$38 indicates more effective turbulent scalar transport than momentum transport, while $r$39 indicates the reverse.

Measurements above three-dimensional urban-canopy arrays found that $r$40 increased with height: $r$41 and

$r$42

The isolated-flow configuration reached approximately $r$43 at upper levels, whereas wake-interference and skimming-flow configurations remained closer to $r$44. These results indicate that $r$45 is a flow property rather than a universal constant (Bernardino et al., 2020).

A height-dependent formulation obtained by combining Prandtl’s momentum diffusivity with a Lagrangian scalar diffusivity is

$r$46

It reproduced the measured vertical trends reasonably for the tested staggered cube arrays. A constant $r$47 remains an engineering approximation, but its appropriate value depends on height, canopy geometry, turbulence structure, source location, flow regime, and calibration conventions. The distinction matters because

$r$48

underestimating $r$49 overestimates scalar mixing and plume dilution, whereas overestimating it underestimates turbulent scalar diffusivity.

7. Extensions and resource-theoretic interpretations

Schmidt-number breaking channels generalize entanglement-breaking channels. A channel $r$50 is $r$51-Schmidt-number breaking if

$r$52

for every bipartite input. This holds if and only if the Choi state satisfies

$r$53

Entanglement-breaking channels are exactly the $r$54 case. For a depolarizing channel,

$r$55

the channel is $r$56-Schmidt-number breaking when

$r$57

This produces the hierarchy

$r$58

Schmidt-number annihilation instead concerns entanglement internal to a composite system $r$59, rather than entanglement between an external reference and the channel output. The two notions coincide only in their respective $r$60 analogues, entanglement breaking and entanglement annihilation (Mallick et al., 2024).

The absolute Schmidt number asks whether a state’s Schmidt number can be increased by any global unitary. A state belongs to $r$61-absolute Schmidt-number-bounded states if

$r$62

This property differs from local-unitary invariance: local unitaries preserve Schmidt number, whereas global unitaries can rearrange eigenvectors and increase entanglement dimensionality. The corresponding nonabsolute states are those for which some global unitary raises the Schmidt number above $r$63 (Mallick et al., 2 Apr 2026).

Absolute Schmidt number admits witness-based and moment-based detection, robustness measures, and channel extensions. For covariant channels, absolute $r$64-Schmidt-number channels coincide with global $r$65-Schmidt-number-annihilating channels. These constructions treat entanglement dimensionality as a resource distinct from ordinary entanglement, distillability, Bell nonlocality, and channel transmissibility.