betaFeedback

WorksPhysics and Astronomy

High-Dimensional Maximally Entangled OAM States by Joint Pump and Modal-Space Engineering

Abstract

Spontaneous parametric down-conversion (SPDC) generates correlated photon pairs that inherently exhibit high-dimensional spatial entanglement in orbital angular momentum (OAM) modes. In practice, the resulting two-photon states are typically not maximally entangled due to a highly non-uniform distribution of generated OAM modes. Although pump-beam engineering has emerged as a tool to prepare high-dimensional entangled states, it has usually been restricted to low-dimensional subspaces, i.e., small sets of selected OAM modes for each photon within the full infinite-dimensional two-photon Hilbert space. A general understanding of the requirements for the generation of maximally entangled states (MESs) for any dimensionality dd with d>2d>2 is lacking. Here, we demonstrate how pump engineering enables the generation of high-dimensional MESs for arbitrary dd defined in d×dd \times d subspaces. By identifying the interplay between the spatial pump structure and the choice of OAM subspace, the required pump profile for a target MES is directly determined. Our results reveal that maximal entanglement with OAM requires a careful and nontrivial design of the effective high-dimensional Hilbert subspace.

Introduction

Photons are a versatile platform for quantum information [1, 2, 3]. High-dimensional entangled states (HDES), where each photon is described by a Hilbert space of dimension d>2d > 2 [4], attract interest for increased information capacity and enhanced robustness in quantum communication [5, 6], reduced circuit complexity in quantum computing [7], and increased phase sensitivity in quantum metrology [8, 9].

Transverse spatial modes of light, particularly orbital angular momentum (OAM) modes, provide a discrete and unbounded basis for high-dimensional photonic entanglement [10, 11, 12, 13, 14]. OAM-based states are used in high-capacity free-space communication [15, 16, 17], dense multiplexing [18, 19], and superresolution imaging [20]. Photon pairs generated via spontaneous parametric down-conversion (SPDC) [21] intrinsically exhibit high-dimensional entanglement in the spatial degree of freedom based on the conservation of OAM [22, 23].

The generated OAM modes are typically non-uniformly distributed, yielding non-maximally entangled two-photon states [24, 25]. However, many quantum-information protocols are optimized for, or explicitly formulated in terms of, maximally entangled states (MESs), the high-dimensional analogues of Bell states [26]. MESs maximize the entanglement entropy, uniformly exploit the full available Hilbert space, and serve as ideal resources to achieve optimal fidelity and error resilience in quantum-information protocols.

Crystal engineering [27, 28, 29] can be applied to shape the OAM spectrum. However, this technique usually requires the fabrication of tailored nonlinear media that are optimized for fixed dimensionality dd. Furthermore, post-generation schemes like entanglement concentration [30, 31, 32, 33] can flatten the measured OAM spectrum, but fail to achieve genuine at-source state engineering and add complexity to the setup.

Furthermore, pump-beam engineering, e.g., via spatial light modulators (SLMs), enables programmable control of the OAM spectrum and has proven to be an experimentally convenient tool for preparing HDES [34, 35, 36]. Yet, the generation of arbitrary-dimensional MESs remains elusive: prior works are often constrained to small-dimensional subspaces [37, 38, 39] or rely on technically demanding optimization procedures of the pump profile [40, 42, 43] [41], without establishing general conditions for maximal entanglement.

In this paper, we introduce a broadly applicable approach for generating MESs in d×dd \times d-dimensional OAM subspaces. Our method jointly engineers the pump beam and local OAM subspaces to establish conditions for maximal entanglement without numerical optimization of the pump profile. Our analytical construction is not restricted to a specific phase-matching type or nonlinear medium. Moreover, the proposed scheme ensures the spectral and spatial purity of the generated state. Thus, our pump-engineering approach provides, together with modal postselection, a systematic construction to generate pure MESs in flexibly chosen OAM subspaces of, in principle, arbitrary dimensionality.

Sketch of the SPDC process and the pump-imprinted OAM spectra

Figure 1. Sketch of the SPDC process and the pump-imprinted OAM spectra. (a) During SPDC, the spatial structure of a pump field formed by a coherent superposition of LG modes with different OAM indices ℓp\ell_p is transferred to the OAM correlations of the generated photon pairs. The two-photon state of signal and idler is a coherent, non-separable superposition of OAM modes ∣ℓ⟩s|\ell\rangle_s and ∣ℓp−ℓ⟩i|\ell_p-\ell\rangle_i, constrained by OAM conservation. (b) The corresponding normalized OAM spectrum of a superposition of pump modes ℓp=0\ell_p=0 (blue) and ℓp=3\ell_p=3 (pink) with the characteristic anti-diagonal correlations in the OAM spectrum. For an equally weighted superposition, the maximum amplitude along the blue diagonal does not match the maximum amplitude along the pink diagonal. (c) If the relative weight of a3a_3 compared to a0a_0 is increased, the maximum amplitudes along the two anti-diagonals can be equalized. (d) An appropriate choice of pump waist wpw_p minimizes the spiral bandwidth, concentrating the population around the two-photon modes ∣0,0⟩|0,0\rangle (blue) or ∣1,2⟩|1,2\rangle and ∣2,1⟩|2,1\rangle (pink).

Theoretical Background

Two-photon entanglement in OAM

Since the pioneering experiment by Mair et al. [22], which demonstrated OAM conservation and entanglement in the SPDC process, many works have studied the controlled generation of high-dimensional entanglement, recently even at the single-photon level [44]. To analyze OAM entanglement in SPDC, the biphoton state is conveniently expanded in an eigenbasis ∣ℓ⟩|\ell\rangle of the OAM operator L^\hat{L}, i.e., L^∣ℓ⟩=ℓℏ∣ℓ⟩\hat{L}|\ell\rangle=\ell\hbar|\ell\rangle with ℓ∈Z\ell\in\mathbb{Z}. Laguerre–Gaussian (LG) modes provide a natural basis for this decomposition [45]. For generality, we assume that the pump beam is also described by an LG mode. Consequently, the entangled state of a signal (s) and idler (i) photon from SPDC driven by an LG pump beam (p) carrying an OAM ℓp\ell_p in the paraxial, quasi-collinear regime (cf. Fig. 1(a)) is

∣ψ⟩=∑ℓsCℓs(ℓp,wp)∣ℓs⟩∣ℓp−ℓs⟩(1)|\psi\rangle=\sum_{\ell_s} C_{\ell_s}(\ell_p,w_p)|\ell_s\rangle|\ell_p-\ell_s\rangle \tag*{(1)}

where OAM conservation sets ℓi=ℓp−ℓs\ell_i=\ell_p-\ell_s. The expansion amplitudes CℓsC_{\ell_s} depend on the experimental geometry, the nonlinear crystal, and the pump beam [46, 47, 48, 49, 50]. Since we aim to shape the pump beam parameters to design HDES, we explicitly emphasize its waist wpw_p and OAM ℓp\ell_p. The OAM conservation directly correlates the signal and idler modes. The projection probability for the two-photon mode ∣ℓs,ℓi⟩|\ell_s,\ell_i\rangle is given by Pℓs,ℓi(ℓp,wp):=∣Cℓs(ℓp,wp)∣2P_{\ell_s,\ell_i}(\ell_p,w_p):=|C_{\ell_s}(\ell_p,w_p)|^2. Here, we consider degenerate SPDC where the wave numbers satisfy kp/2=ks=kik_p/2=k_s=k_i. This, combined with equal projection waists for signal and idler, enforces the symmetry condition Cℓs=Cℓp−ℓsC_{\ell_s}=C_{\ell_p-\ell_s} [29]. In addition to the azimuthal index ℓ\ell, Laguerre–Gaussian modes are

characterized by a radial index pp [51]. Different OAM components can be correlated with distinct radial-mode states. Tracing over unresolved radial degrees of freedom can therefore reduce the purity of the OAM state. Our main results are formulated for signal and idler modes selected at zeroth radial order, ps=pi=0p_s=p_i=0. The detailed description of the state in the broadband regime can be found in Appendix A and higher-order radial correlations [51, 52, 53, 54, 55] in Appendix B. The same analytical framework and engineering procedure also apply in the broadband regime, where the pump weights are determined from frequency-integrated coincidence probabilities. Moreover, the target mode pairs in our construction share the same spectral dependence under the assumptions of our model. The conditional target state therefore retains its spatial purity after tracing over frequency [56] as described in Appendix § 0.3.3.

Pump-beam engineering

Central to our approach is the freedom to utilize a general spiral distribution for the pump. Mathematically, this is expressed as an arbitrary superposition of LG modes [29]

Vp=∑ℓpaℓp LG(ℓp,wp).V_p=\sum_{\ell_p}a_{\ell_p}\,\mathrm{LG}(\ell_p,w_p).

with ∑ℓp∣aℓp∣2=1\sum_{\ell_p}|a_{\ell_p}|^2=1. Such a pump profile can be readily realized using commercially available SLMs [57]. In the low-gain regime of SPDC, the generated two-photon amplitude is linear in the pump field [56]. Therefore, the two-photon state generated by a pump superposition is the coherent sum of the contributions associated with

@page 2/15 starts_mid=0 ends_mid=1

Sketch of the SPDC process and the pump-imprinted OAM spectra

Figure 1 (PDF p. 2). Sketch of the SPDC process and the pump-imprinted OAM spectra. (a) During SPDC, the spatial structure of a pump field formed by a coherent superposition of LG modes with different OAM indices ℓp\ell_p is transferred to the OAM correlations of the generated photon pairs. The two-photon state of signal and idler is a coherent, non-separable superposition of OAM modes ∣ℓ⟩s|\ell\rangle_s and ∣ℓp−ℓ⟩i|\ell_p-\ell\rangle_i, constrained by OAM conservation. (b) The corresponding normalized OAM spectrum of a superposition of pump modes ℓp=0\ell_p=0 (blue) and ℓp=3\ell_p=3 (pink) with the characteristic anti-diagonal correlations in the OAM spectrum. For an equally weighted superposition, the maximum amplitude along the blue diagonal does not match the maximum amplitude along the pink diagonal. (c) If the relative weight of a3a_3 compared to a0a_0 is increased, the maximum amplitudes along the two anti-diagonals can be equalized. (d) An appropriate choice of pump waist wpw_p minimizes the spiral bandwidth, concentrating the population around the two-photon modes ∣0,0⟩|0,0\rangle (blue) or ∣1,2⟩|1,2\rangle and ∣2,1⟩|2,1\rangle (pink).

Theoretical Background

Two-photon entanglement in OAM

Since the pioneering experiment by Mair et al. [22], which demonstrated OAM conservation and entanglement in the SPDC process, many works have studied the controlled generation of high-dimensional entanglement, recently even at the single-photon level [44]. To analyze OAM entanglement in SPDC, the biphoton state is conveniently expanded in an eigenbasis ∣ℓ⟩|\ell\rangle of the OAM operator L^\hat{L}, i.e., L^∣ℓ⟩=ℓℏ∣ℓ⟩\hat{L}|\ell\rangle=\ell\hbar|\ell\rangle with ℓ∈Z\ell\in\mathbb{Z}. Laguerre–Gaussian (LG) modes provide a natural basis for this decomposition [45]. For generality, we assume that the pump beam is also described by an LG mode. Consequently, the entangled state of a signal (s) and idler (i) photon from SPDC driven by an LG pump beam (p) carrying an OAM ℓp\ell_p in the paraxial, quasi-collinear regime (cf. Fig. 1(a)) is

∣ψ⟩=∑ℓsCℓs(ℓp,wp)∣ℓs⟩∣ℓp−ℓs⟩(2)|\psi\rangle=\sum_{\ell_s} C_{\ell_s}(\ell_p,w_p)|\ell_s\rangle|\ell_p-\ell_s\rangle \tag*{(2)}

where OAM conservation sets ℓi=ℓp−ℓs\ell_i=\ell_p-\ell_s. The expansion amplitudes CℓsC_{\ell_s} depend on the experimental geometry, the nonlinear crystal, and the pump beam [46, 47, 48, 49, 50]. Since we aim to shape the pump beam parameters to design HDES, we explicitly emphasize its waist wpw_p and OAM ℓp\ell_p. The OAM conservation directly correlates the signal and idler modes. The projection probability for the two-photon mode ∣ℓs,ℓi⟩|\ell_s,\ell_i\rangle is given by Pℓs,ℓi(ℓp,wp):=∣Cℓs(ℓp,wp)∣2P_{\ell_s,\ell_i}(\ell_p,w_p):=|C_{\ell_s}(\ell_p,w_p)|^2. Here, we consider degenerate SPDC where the wave numbers satisfy kp/2=ks=kik_p/2=k_s=k_i. This, combined with equal projection waists for signal and idler, enforces the symmetry condition Cℓs=Cℓp−ℓsC_{\ell_s}=C_{\ell_p-\ell_s} [29]. In addition to the azimuthal index ℓ\ell, Laguerre–Gaussian modes are

characterized by a radial index pp [51]. Different OAM components can be correlated with distinct radial-mode states. Tracing over unresolved radial degrees of freedom can therefore reduce the purity of the OAM state. Our main results are formulated for signal and idler modes selected at zeroth radial order, ps=pi=0p_s=p_i=0. The detailed description of the state in the broadband regime can be found in Appendix A and higher-order radial correlations [51, 52, 53, 54, 55] in Appendix B. The same analytical framework and engineering procedure also apply in the broadband regime, where the pump weights are determined from frequency-integrated coincidence probabilities. Moreover, the target mode pairs in our construction share the same spectral dependence under the assumptions of our model. The conditional target state therefore retains its spatial purity after tracing over frequency [56] as described in Appendix § 0.3.3.

Pump-beam engineering

Central to our approach is the freedom to utilize a general spiral distribution for the pump. Mathematically, this is expressed as an arbitrary superposition of LG modes [29]

Vp=∑ℓpaℓp LG(ℓp,wp).V_p=\sum_{\ell_p}a_{\ell_p}\,\mathrm{LG}(\ell_p,w_p).

with ∑ℓp∣aℓp∣2=1\sum_{\ell_p}|a_{\ell_p}|^2=1. Such a pump profile can be readily realized using commercially available SLMs [57]. In the low-gain regime of SPDC, the generated two-photon amplitude is linear in the pump field [56]. Therefore, the two-photon state generated by a pump superposition is the coherent sum of the contributions associated with each pump OAM component,

∣ψ~⟩=∑ℓpaℓp∑ℓsCℓs(ℓp,wp)∣ℓs,ℓi⟩.(3)|\tilde{\psi}\rangle= \sum_{\ell_p} a_{\ell_p} \sum_{\ell_s} C_{\ell_s}(\ell_p,w_p) |\ell_s,\ell_i\rangle. \tag*{(3)}

Each pump component with ℓp\ell_p populates one anti-diagonal in the OAM spectrum, the distribution of Pℓs,ℓiP_{\ell_s,\ell_i}, while the coefficients aℓpa_{\ell_p} control the relative weights between different anti-diagonals. Fig. 1 (b) shows the normalized OAM spectrum for an equally weighted superposition of the pump modes LG(0,wp)(0,w_p) and LG(3,wp)(3,w_p). The resulting spectrum consists of two anti-diagonals, shown in blue and pink, each satisfying OAM conservation for the corresponding pump component. In contrast to the single-pump scenario, the signal and idler OAM modes are no longer connected by a single anti-correlation rule: a given signal mode can be paired with different idler modes belonging to different pump anti-diagonals.

Equal pump weights, a0=a3a_0=a_3, do not generally produce equal two-photon amplitudes on the two anti-diagonals. The pump weights aℓpa_{\ell_p} can therefore be adjusted to rescale the contributions of the different anti-diagonals. To equalize different OAM amplitudes, the calculation of the pump weights can be easily obtained via a matrix equation for the state coefficients (see Appendix § 0.3.3). For example, the coefficient a0=(d C0(0,wp))−1a_0=(\sqrt{d}\,C_0(0,w_p))^{-1} enforces an amplitude 1/d1/\sqrt{d} for the state ∣0,0⟩|0,0\rangle. In Fig. 1 (c), the weights are chosen such that the maxima of each anti-diagonal are equalized.

Each anti-diagonal represents an independent distribution of possible modes, characterized by a width, called spiral bandwidth [58]. The spiral bandwidth can be reduced by tuning the pump waist, concentrating the distribution near its dominant amplitudes [59], as shown in Fig. 1 (d). The dominant amplitudes occur at the center of this anti-diagonal. Usually, this setting is referred to as the Single-Schmidt mode regime (SSMR). Here, higher-radial orders are minimized too [38, 59]. For even ℓp\ell_p, the strongest contribution is the symmetric mode pair with OAM ∣ℓp2,ℓp2⟩|\frac{\ell_p}{2},\frac{\ell_p}{2}\rangle. For odd ℓp\ell_p, the center lies between two OAM modes, so the two neighboring central amplitudes of the modes ∣ℓp+12,ℓp−12⟩|\frac{\ell_p+1}{2},\frac{\ell_p-1}{2}\rangle and ∣ℓp−12,ℓp+12⟩|\frac{\ell_p-1}{2},\frac{\ell_p+1}{2}\rangle are degenerate and maximal as shown in Fig. 1 (b) and (c). For ∣ℓp∣≥2|\ell_p|\ge2, the finite spiral bandwidth gives rise to additional neighboring modes with non-negligible amplitudes, which cannot be completely suppressed by pump shaping alone [60]. Thus, while ℓp=0\ell_p=0 yields a strongly populated mode ∣0,0⟩|0,0\rangle, higher pump modes produce broader spectra with non-vanishing contributions, as evident from Fig. 1 (c).

Subspaces and maximal entanglement

Although the OAM Hilbert space is formally infinite, the number of significantly populated modes in a practical setup is finite. We restrict the state preparation and measurement to selected dd-dimensional OAM subspaces. Treating the signal and idler photons as qudits, we define the dd-dimensional single-photon subspace Sd=span⁡{∣ℓj⟩},j=1,…,dS_d=\operatorname{span}\{|\ell_j\rangle\},j=1,\ldots,d. We further distinguish between continuous OAM subspaces, with modes in a continuous interval of OAM mode numbers, and subspaces in a discontinuous range of OAM mode numbers, here called disassembled OAM subspaces. For the biphotonic system, the associated two-photon subspace is Sd×d=Sd⊗SdS_{d\times d}=S_d\otimes S_d of dimension d2d^2.

By definition, a maximally entangled state in this subspace is a pure bipartite state for which tracing over either subsystem yields a maximally mixed state for the other [26]. Here, we quantify the entanglement within the two-photon subspace Sd×dS_{d\times d} using the Schmidt decomposition [61, 62] and the Schmidt number K\mathcal{K} bounds the entanglement, ranging from K=1\mathcal{K}=1 for a separable state to K=d\mathcal{K}=d for an MES [29]. For the states described by Eqs. (1) and (3), an MES in Sd×dS_{d\times d} is formed by a coherent superposition of dd orthogonal mode pairs, each with an amplitude of 1/d1/\sqrt{d}, while the remaining d2−dd^2-d mode amplitudes vanish. The condition for maximal entanglement K=d\mathcal{K}=d is satisfied only if both these requirements are met. Definitions based solely on equal amplitudes [38, 63] are incomplete, as such states can result in reduced states that are not maximally mixed. In the OAM basis, the two conditions correspond to exactly one populated two-photon mode in each row and each column of the d×dd\times d subspace, with all dd amplitudes being equal. As this logic is very similar to the rule of filling a number in a Sudoku puzzle, we refer to this as the Sudoku condition. The rest of the modes within the subspace must have negligibly small amplitudes. Since local unitary operations relate all MES within a given OAM subspace, it is sufficient to concentrate on a single MES for each dimension dd.

Any population outside the target subspace decreases the effective photon-pair rate, as these components are projected out when measuring solely within the selected OAM modes. These contributions outside of the considered subspace are rarely thoroughly quantified in most works. Hence, we introduce the global fidelity Fglob=⟨ψtar∣ψ⟩\mathcal{F}_{\mathrm{glob}}=\langle\psi_{\mathrm{tar}}|\psi\rangle that measures the overlap of the full generated state ∣ψ⟩|\psi\rangle and the maximally entangled target state ∣ψtar⟩∈Sd×d|\psi_{\mathrm{tar}}\rangle\in S_{d\times d}. A value of Fglob=1\mathcal{F}_{\mathrm{glob}}=1 indicates that the entire generated OAM spectrum perfectly matches the target state. We also define a subspace fidelity, where the full state ∣ψ⟩|\psi\rangle is restricted to the subspace modes. Thus, Fsub=1\mathcal{F}_{\mathrm{sub}}=1 corresponds to a maximally entangled state within Sd×dS_{d\times d}.

We note that we will search by design for conditional states that are only maximally entangled after projection onto a selected OAM subspace for a given dimensionality. If the amplitudes of the selected target modes are not the dominant amplitudes of the full generated spectrum, the state preparation can be globally very inefficient and should be discarded. Figure 6 and 7 in Appendix § 0.3.3 illustrate such cases.

OAM spectra pumped with superposition of OAM indices $\ell_p=0,3$ and different radial indices

Figure 6. OAM spectra pumped with superposition of OAM indices ℓp=0,3\ell_p=0,3, but different radial indices of the pump beam compared. (a) Fig. 1 with pp=0p_p=0 reconstructed. Within the subspace we find K=3\mathcal{K}=3, that corresponds to Fsub=1\mathcal{F}_{\mathrm{sub}}=1 for the target MES. The global fidelity is Fglob=0.68\mathcal{F}_{\mathrm{glob}}=0.68. (b) For pp=1p_p=1, we locally generate the same MES, optimized in the subspace. However, the fidelity is practically zero since these are only small contributions to the global state. This illustrates that subspace optimization is only efficient using the maxima of the anti-diagonals as target modes.

Results

A coherent superposition of appropriately weighted pump OAM modes provides a direct tool for satisfying the Sudoku condition within a selected dd-dimensional OAM subspace. Here, we investigate the relationship between the choice of target modes for signal and idler, the structure of the local OAM subspaces, and the OAM superposition of the pump beam.

Dimensionality limit in continuous subspaces

We first consider continuous subspaces of dimension dd and show that our construction of subspace-optimized MES with K≈d\mathcal{K}\approx d can only be realized up to d=6d=6. As discussed above, depending on the parity of ℓp\ell_p, we can have one central mode (even ℓp\ell_p) or two central modes (odd ℓp\ell_p) with maximum amplitude per anti-diagonal. We can recognize three different configurations to realize an MES for d≤6d\leq6.

The first configuration is to populate modes along the diagonal ℓs=ℓi\ell_s=\ell_i as shown in Figure 2 (a) yielding the MES 13(∣−1,−1⟩+∣0,0⟩+∣1,1⟩)\frac{1}{\sqrt{3}}(|-1,-1\rangle+|0,0\rangle+|1,1\rangle) in the subspace S3×3\mathcal{S}_{3\times3} defined by S3=span⁡{∣−1⟩,∣0⟩,∣1⟩}\mathcal{S}_3=\operatorname{span}\{|-1\rangle,|0\rangle,|1\rangle\}. As a direct consequence of the non-vanishing modes adjacent to the target modes, we have Fglob<1\mathcal{F}_{\mathrm{glob}}<1. This method fails for larger d>3d>3: for example in S5=span⁡{∣−2⟩,…,∣2⟩}\mathcal{S}_5=\operatorname{span}\{|-2\rangle,\ldots,|2\rangle\}, OAM conservation introduces additional contributions from a±2a_{\pm2} that reduce the Schmidt number, and the state is no longer maximally entangled, see Figure 2 (b). The global fidelity in Figure 2 (b) is further reduced compared to Figure 2 (a) due to the increased number of terms in the pump superposition.

Different methods to engineer MES within continuous subspaces

Figure 2. Different methods to engineer MES within continuous subspaces. The fidelities Fsub\mathcal{F}_{\mathrm{sub}} and Fglob\mathcal{F}_{\mathrm{glob}} quantify the overlap with the target MES within the selected subspace, marked in red, and within the full LG mode space, respectively. Populating the diagonal ℓs=ℓi\ell_s=\ell_i works for d=3d=3 in (a), but fails for larger dimensions due to non-vanishing mode contributions within the target subspace, as shown for d=5d=5 in (b). Alternatively, by placing modes at or near the subspace boundary in the upper left or lower right in (c) and (d), unwanted contributions are suppressed. A hybrid method is shown in (e) and (f) by populating modes at the center of the subspace and at the boundary. Scaling to higher dimensionality dd requires equalizing modes neighboring the center modes, as indicated by the yellow dashed boxes in (g), which cannot effectively be done with pump beam shaping.

For d=3d=3 and d=4d=4, a boundary-maxima configuration can be used, in which the central maxima of the selected anti-diagonals are placed near the boundaries of the d×dd\times d OAM subspace. This arrangement shifts most unwanted neighboring contributions outside the selected mode space. For S3\mathcal{S}_3, the pump components a−1a_{-1} and a2a_2 yield the MES 13(∣−1,0⟩+∣0,−1⟩+∣1,1⟩)\frac{1}{\sqrt{3}}(|-1,0\rangle+|0,-1\rangle+|1,1\rangle), which achieves a higher global fidelity than the diagonal configuration, see Figure 2 (c). The same principle extends to d=4d=4, as shown in Figure 2 (d).

For d=5d=5 and d=6d=6, the boundary configuration alone is no longer sufficient. Instead, degenerate modes near the boundary must be combined with modes in the subspace center. Such center modes can only be prepared with pump weights a0a_0 and a±1a_{\pm1}, since all other pump modes introduce undesired, adjacent modes inside the subspace. This enables the only viable MES in S5×5\mathcal{S}_{5\times5} shown in Figure 2 (e). For continuous S6×6\mathcal{S}_{6\times6} subspaces we can use a±1a_{\pm1} for the center modes. An example is shown in Figure 2 (f) with S6=span⁡{∣−2⟩,…,∣3⟩}\mathcal{S}_6=\operatorname{span}\{|-2\rangle,\ldots,|3\rangle\}.

For d≥7d\geq7, however, continuous subspaces will fail: exactly matching only a few required amplitudes to those of the center modes, like the two modes indicated by the yellow dashed square in Figure 2 (g), cannot be efficiently achieved by tuning the pump alone [60].

Appendix D discusses matching different amplitudes along the same anti-diagonal ℓp=0\ell_p = 0 for large pump beam waist (approaching the plane-wave-limit). Including higher-order radial pump modes could enable equal amplitudes of such modes as well, but can significantly reduce the global fidelity due to increased radial correlations. For example, Fig. 7 (b) in the Appendix shows a MES for d=7d = 7 with Schmidt number K=6.91K = 6.91. Preparing this state, however, comes at the cost of a lower global fidelity and the loss of the analytical ease of calculating the pump spectrum. This motivates the use of disassembled subspaces.

MES preparation based on disassembled subspaces

We first illustrate the idea of a disassembled subspace with a modified configuration from the continuous subspace of Figure 2 (e): removing the pump component a0a_0 gives a spectrum shown in Figure 3 (a). The remaining pump weights a±3a_{\pm3} generate four degenerate central amplitudes. Instead of keeping the continuous interval from before, we select only the OAM modes that support these amplitudes, defining the subspace S4=span⁡{∣−2⟩,∣−1⟩,∣1⟩,∣2⟩}S_4 = \operatorname{span}\{\lvert-2\rangle, \lvert-1\rangle, \lvert1\rangle, \lvert2\rangle\}. Within this disassembled subspace, the generated state ∣ψ⟩=12(∣−2,−1⟩+∣−1,−2⟩+∣1,2⟩+∣2,1⟩)\lvert\psi\rangle= \frac{1}{2}(\lvert-2,-1\rangle+ \lvert-1,-2\rangle+ \lvert1,2\rangle+ \lvert2,1\rangle) is a four-dimensional MES. This example demonstrates the basic role of disassembled subspaces: modes that are not needed for the target state

Pump engineering scheme in disassembled subspaces

Figure 3. Pump engineering scheme in disassembled subspaces. (a) The range of the available modes in a disassembled subspace is not continuous. The choice of pump modes determines which signal and idler modes can be selected by picking the maxima along the anti-diagonals. Compared to Figure 2 (e), we omit a0a_0 in the pump superposition, and the remaining pump modes generate four degenerate central amplitudes. We select only the OAM modes that support these amplitudes, which defines the disassembled subspace S4S_4. (b) Not all pump configurations preserve maximal entanglement: a±5a_{\pm5} introduces separable modes within the subspace S6S_6, reducing the subspace fidelity. The state is not maximally entangled. (c) A sufficient difference in the OAM indices of the pump ensures well-separated two-photon modes that can form MESs. When the subspace is extended, e.g., by including modes ∣±3⟩\lvert\pm3\rangle and ∣±4⟩\lvert\pm4\rangle, the appropriate choice of pump modes a±7a_{\pm7} enables an MES for d=8d = 8 and excludes unwanted contributions.

are excluded, so that unwanted contributions do not enter the effective d×dd \times d Hilbert space. The global fidelity is Fglob=0.62\mathcal{F}_{\mathrm{glob}} = 0.62, slightly lower than in the continuous-subspace example (Figure 2 (d)), as our choice of pump modes here increases the number of undesired modes.

Next, we show that this approach can be extended beyond the dimensionality limit (with d>6d > 6) for MESs in a continuous subspace. As an example, we consider the dimension d=8d = 8. The subspace from Figure 3 (a) can be extended stepwise by adding suitable pump terms. A natural first attempt is to use four odd pump components and equalize their central amplitudes. Yet, amplitude equalization alone is not sufficient. As shown in Figure 3 (b), choosing pump modes that are too close to each other (a±5a_{\pm5}) can produce eight equally weighted amplitudes, but the Sudoku condition for an eight-dimensional MES is violated. Consequently, the corresponding mode pairs are not independent, and signal and idler modes are reused within the subspace. The subspace modes form a 6×66 \times6 subspace, but the state is not maximally entangled, K=4≠6K = 4 \ne6.

This problem is avoided by increasing the separation between the anti-diagonals, i.e., the difference in pump OAM indices. In particular, replacing the additional pump components by a±7a_{\pm7} yields new central mode pairs that do not overlap with those generated by a±3a_{\pm3}. The resulting target modes define the disassembled subspace S8=span⁡{∣−4⟩,∣−3⟩,∣−2⟩,∣−1⟩,∣1⟩,∣2⟩,∣3⟩,∣4⟩}S_8 = \operatorname{span}\{\lvert-4\rangle, \lvert-3\rangle, \lvert-2\rangle, \lvert-1\rangle, \lvert1\rangle, \lvert2\rangle, \lvert3\rangle, \lvert4\rangle\}, within which the eight equal amplitudes form a eight-dimensional MES, cf. Figure 3 (c).

This illustrates the central principle of the approach: the OAM subspace and the choice of the OAM pump mode weights are closely related. High-dimensional MESs can be constructed by introducing sufficient separation between the pump modes, equalizing the desired central amplitudes through their relative weights, and selecting a subspace that retains these target modes while excluding unwanted neighboring contributions. In general, several valid pump-subspace combinations may exist for a given dimension.

When comparing Fig. 1 (d) to Figure 2 (a) and (c), we recognize that multiple possibilities for d=3d = 3 exist to choose pump modes or subspace modes. Higher OAM indices for the pump modes can equivalently enable conditioned maximally entangled qutrits, but the number of non-vanishing modes adjacent to the produced target modes increases. The same holds for the construction of MESs in disassembled states. The following design criteria for suitable MESs and subspaces can be summarized as:

  1. Fulfill the Sudoku condition (one target mode per column and row, all with equal amplitudes). The OAM spectrum in the conditioned subspace represents a d×dd \times d MES with Schmidt number K≈dK \approx d. In the broadband regime, we have to make sure that all target modes have the same spectral dependence to ensure spatial purity.

  1. The amplitudes within the subspace are the global maxima of all amplitudes. Prevent pump modes that generate the most probable modes for signal and idler outside the target subspace.

  1. Maximize the global fidelity via waist optimization to enhance the population within the target subspace and to reduce mode leakage contributions outside the subspace.

To maximize the global fidelity, one should minimize both the number of pump components and the spiral bandwidth of each component. The latter can be achieved in the SSMR. Odd pump OAM modes are particularly useful because each of them contributes two degenerate target mode pairs, enabling larger dimensionalities with fewer pump modes.

Figure 4 (a) shows how our approach can be readily extended to larger dimensions. For example, we make use of a±15a_{\pm15} and find a 12×1212 \times12 MES if subspace modes {∣±7⟩,∣±8⟩}\{\lvert\pm7\rangle, \lvert\pm8\rangle\} are included, and {∣±5⟩,∣±6⟩}\{\lvert\pm5\rangle, \lvert\pm6\rangle\} are excluded. A further increase in dimensionality requires even larger separations than that between ±7\pm7 and ±15\pm15.

Joint pump-beam and modal-space engineering based on disassembled subspaces extended to higher dimensions

Figure 4. Joint pump-beam and modal space engineering based on disassembled subspaces extended to higher dimensions. (a) Six sufficiently separated pump OAM components are used to construct a d=12d = 12 MES. Compared to Figure 3 (c), the subspace requires modes with two additional gaps within the selected subspace range. (b) A summary of the subspace and global fidelities obtained with our approach as functions of the MES dimensionality up to d=12d = 12.

Larger ℓp\ell_p results in a wider bandwidth with non-negligible modes and therefore smaller global fidelities. Therefore, the general trend is that increasing dd of the MES results in a decrease in global fidelity. Figure 4 (b) compares the subspace and global fidelities obtained with our approach as functions of the dimensionality up to d=12d = 12. More details on how the figure is obtained, and which pump and subspace modes are chosen, can be found in Appendix E.

Karan et al. [42] introduced a postselection-free approach based on higher-order radial pump modes, enabling the engineering of approximately uniform OAM spectra while leaving the radial degree of freedom unresolved. Although this approach avoids the losses associated with selecting only the lowest radial order, equalized OAM populations do not necessarily imply a pure maximally entangled OAM state: correlations with unresolved radial or spectral degrees of freedom can reduce coherence between OAM components. In our approach, purity is an explicit design requirement. We select a defined LG-mode subspace whose target mode pairs share the same spectral dependence, ensuring spatial–spectral separability of the target MES. A possible extension of our approach to include higher radial modes is discussed in Appendix F.

Conclusion

In this paper, we introduced a universal pump engineering approach for the design of high-dimensional MESs via SPDC. Our analysis demonstrates that the

joint optimization of the spatial structure of the pump and the careful design of the effective OAM subspace enable the preparation of MESs. For continuous OAM subspaces, the finite spiral bandwidth imposes a dimensionality limit of d≤6d \le6, since the construction of higher-dimensional subspaces inevitably leads to undesired, non-negligible mode contributions within the target space. To overcome this limitation, we introduced the concept of disassembled OAM subspaces. By engineering the OAM pump modes and selecting subspace modes, this approach isolates the desired maximal amplitudes while systematically eliminating unwanted contributions arising from the spiral bandwidth. Ultimately, our combined pump–modal space engineering provides a practical yet extendable approach to generating high-dimensional MESs with, in principle, arbitrary dimensionality.

Details on the biphoton state

SPDC, as a second-order nonlinear process, is inherently probabilistic. We assume a classical, undepleted pump beam driving the SPDC process in the low-gain regime. Here, one pump photon splits into a pair of signal and idler photons when interacting with a second-order nonlinear material. The photons propagate quasi-collinear along kj=qj+kj,zez\mathbf{k}_j = \mathbf{q}_j + k_{j,z}\mathbf{e}_z with ∣q∣≪∣k∣:=k|\mathbf{q}| \ll|\mathbf{k}| := k for j=p,s,ij = \mathrm{p},\mathrm{s},\mathrm{i}, respectively. In the paraxial regime, the down-converted photons of frequency ωj\omega_j, transverse momentum qj\mathbf{q}_j, and polarization σj\sigma_j are generated by creation operators a^†\hat{a}^{\dagger} acting on the vacuum states as

∣ψ⟩∝∑σs,σi∫d2qs dωs d2qi dωi Φ(qs,ωs,qi,ωi)(a^σs†(qs,ωs)⊗a^σi†(qi,ωi))∣vac⟩.|\psi\rangle\propto\sum_{\sigma_s,\sigma_i}\int d^2q_s\,d\omega_s\,d^2q_i\,d\omega_i\,\Phi(\mathbf{q}_s,\omega_s,\mathbf{q}_i,\omega_i) \left(\hat{a}^{\dagger}_{\sigma_s}(\mathbf{q}_s,\omega_s)\otimes\hat{a}^{\dagger}_{\sigma_i}(\mathbf{q}_i,\omega_i)\right)|\mathrm{vac}\rangle.

The biphoton mode function Φ(qs,ωs,qi,ωi)\Phi(\mathbf{q}_s,\omega_s,\mathbf{q}_i,\omega_i) reads

Φ(qs,qi,ωs,ωi)=Vp(qs+qi)Sp(ωs+ωi)∫−L/2L/2dz χ(2)(z)ei(kp,z(qs+qi)−ks,z(qs)−ki,z(qi))z.\Phi(\mathbf{q}_s,\mathbf{q}_i,\omega_s,\omega_i)=V_p(\mathbf{q}_s+\mathbf{q}_i)S_p(\omega_s+\omega_i) \int_{-L/2}^{L/2} dz\,\chi^{(2)}(z)e^{i(k_{p,z}(\mathbf{q}_s+\mathbf{q}_i)-k_{s,z}(\mathbf{q}_s)-k_{i,z}(\mathbf{q}_i))z}.

where the variable nonlinear susceptibility profile χ(2)(z)\chi^{(2)}(z) represents either a periodically poled crystal [49] or a custom-poled structure [50]. VpV_p and SpS_p are the spatial and the spectral distribution of the pump beam, respectively. The biphoton mode function encodes the spatial and spectral properties of the photon pairs and is, in general, non-separable. Transverse momentum conservation qp=qs+qi\mathbf{q}_p=\mathbf{q}_s+\mathbf{q}_i is justified when the transverse crystal dimensions are much larger than the beam waists. There are no restrictions to the longitudinal direction, i.e., the length LL of the crystal. The phase-matching conditions determine the propagation directions of the photons, the allowed frequency range, and the permitted polarization

combinations. We can omit the polarization index σj\sigma_j implicitly by specifying the phase matching as type-0, type-I, or type-II SPDC.

An important aspect of SPDC-based photon sources is the spectral-spatial coupling. Since we are primarily interested in the spatial domain (transverse momentum space), the narrowband (or quasi-monochromatic) regime, achieved experimentally using narrowband spectral filters, effectively projects the signal and idler onto the central frequencies ωs,0\omega_{s,0} and ωi,0\omega_{i,0} satisfying ωp,0=ωs,0+ωi,0\omega_{p,0}=\omega_{s,0}+\omega_{i,0} [64]. Consequently, all subsequent expressions are evaluated at these fixed frequencies. For a broadband spectral description we use ω=ω0+Ω\omega=\omega_0+\Omega, where the frequency deviation Ω\Omega is much smaller than the central frequency Ω≪ω0\Omega\ll\omega_0. Then the Fresnel approximation yields

kz=k(Ω)1−∣q∣2k(Ω)2=k(ω0)+Ωu(ω0)+G(ω0)Ω22−∣q∣22k(ω0).k_z=k(\Omega)\sqrt{1-\frac{|\mathbf{q}|^2}{k(\Omega)^2}}=k(\omega_0)+\frac{\Omega}{u(\omega_0)}+\frac{G(\omega_0)\Omega^2}{2}-\frac{|\mathbf{q}|^2}{2k(\omega_0)}.

with the group velocity uu and the group velocity dispersion GG. For a phase-matched crystal satisfying kp−ks−ki=0k_p-k_s-k_i=0, the remaining phase mismatch is given by

Δkz=∣qs∣22ks(ωs)+∣qi∣22ki(ωi)−∣qs+qi∣22kp(ωp)+ΔΩ\Delta k_z=\frac{|\mathbf{q}_s|^2}{2k_s(\omega_s)}+\frac{|\mathbf{q}_i|^2}{2k_i(\omega_i)}-\frac{|\mathbf{q}_s+\mathbf{q}_i|^2}{2k_p(\omega_p)}+\Delta_\Omega

with

ΔΩ=Ωs+Ωiuq,p−Ωsuq,s−Ωiuq,i+12(Gp(Ωs+Ωi)2−GsΩs2−GiΩi2)\Delta_\Omega=\frac{\Omega_s+\Omega_i}{u_{q,p}}-\frac{\Omega_s}{u_{q,s}}-\frac{\Omega_i}{u_{q,i}}+\frac{1}{2}\left(G_p(\Omega_s+\Omega_i)^2-G_s\Omega_s^2-G_i\Omega_i^2\right)

only being present in the broadband regime. Unless stated otherwise, our analysis uses the following baseline parameters: a 405 nm continuous-wave (cw) laser pumping a 15 mm ppKTP crystal for type-II phase matching, producing photons spectrally centered at 810 nm. Importantly, our method is independent of these specific parameters and is readily adaptable to other configurations.

Expansion amplitude of the full LG basis

Transverse LG modes also carry the radial index pp in addition to the OAM index ℓ\ell. Due to the conservation of OAM, the azimuthal index ℓ\ell plays a central role in the transverse spatial description of SPDC and remains the primary focus of this work. Consequently, we adopted a simplified notation for the LG modes in the main text, omitting the radial index pp. Here, we introduce a notation that captures the complete transverse description of the two-photon state and explain how it relates to the notation used in the main text. When calculating the global fidelity of the full state, the radial description cannot be neglected. The general LG decomposition of the biphoton state is

∣ψ⟩=∑ps,ℓs∑pi,ℓiCps,piℓs,ℓi(∣ps,ℓs⟩s⊗∣pi,ℓi⟩i).(4)|\psi\rangle=\sum_{p_s,\ell_s}\sum_{p_i,\ell_i}C^{\ell_s,\ell_i}_{p_s,p_i}\left(|p_s,\ell_s\rangle_s\otimes|p_i,\ell_i\rangle_i\right). \tag*{(4)}

where the spatial part is expanded into the complete set of orthogonal LG modes ∣p,ℓ⟩=∫dq LG(p,ℓ,w)a^†(q)∣vac⟩|p,\ell\rangle=\int d\boldsymbol{q}\,\mathrm{LG}(p,\ell,w)\hat{a}^{\dagger}(\boldsymbol{q})|\mathrm{vac}\rangle, that are eigenstates of the OAM operator L^∣p,ℓ⟩=ℓℏ∣p,ℓ⟩\hat{L}|p,\ell\rangle=\ell\hbar|p,\ell\rangle, ℓ∈Z\ell\in\mathbb{Z} and where ww is the waist of the LG mode. The full expression for the LG modes in momentum space is

LGpℓ(q,w)=w2p!2π(p+∣ℓ∣)!(−1)p+ℓ2(w∣q∣2)∣ℓ∣e−∣q∣2w24Lp∣ℓ∣(∣q∣2w22)eiℓArg⁡(q).\mathrm{LG}_{p}^{\ell}(\boldsymbol{q},w)=\sqrt{\frac{w^{2}p!}{2\pi(p+|\ell|)!}}(-1)^{p+\frac{\ell}{2}}\left(\frac{w|\boldsymbol{q}|}{\sqrt{2}}\right)^{|\ell|}e^{-\frac{|\boldsymbol{q}|^{2}w^{2}}{4}}L_{p}^{|\ell|}\left(\frac{|\boldsymbol{q}|^{2}w^{2}}{2}\right)e^{i\ell\operatorname{Arg}(\boldsymbol{q})}.
Cps,piℓs,ℓi(pp,ℓp,wp;χ(2),L;ws,wi)=⟨ps,ℓs∣⊗⟨pi,ℓi∣ψ⟩=N∫−L2L2dz∬dqs dqi χ(2)(z)eiΔkz(qs,qi)zLGppℓp(qs+qi,wp)[LGpsℓs(qs,ws)]∗[LGpiℓi(qi,wi)]∗∝δℓs+ℓi,ℓp.\begin{aligned} C_{p_s,p_i}^{\ell_s,\ell_i}\left(p_p,\ell_p,w_p;\chi^{(2)},L;w_s,w_i\right)=\langle p_s,\ell_s|\otimes\langle p_i,\ell_i|\psi\rangle\\ &=\mathcal{N}\int_{-\frac{L}{2}}^{\frac{L}{2}}dz\iint d\boldsymbol{q}_s\,d\boldsymbol{q}_i\,\chi^{(2)}(z)e^{i\Delta k_z(\boldsymbol{q}_s,\boldsymbol{q}_i)z}\mathrm{LG}_{p_p}^{\ell_p}(\boldsymbol{q}_s+\boldsymbol{q}_i,w_p)\left[\mathrm{LG}_{p_s}^{\ell_s}(\boldsymbol{q}_s,w_s)\right]^*\left[\mathrm{LG}_{p_i}^{\ell_i}(\boldsymbol{q}_i,w_i)\right]^*\propto\delta_{\ell_s+\ell_i,\ell_p}. \end{aligned}

The collection waists for signal and idler ws,wiw_s,w_i are determined by the experimental setup and projective detection scheme. The Kronecker delta expresses the well-known OAM conservation and the resulting OAM anti-correlations between signal and idler photons ℓi=ℓp−ℓs\ell_i=\ell_p-\ell_s.

This description of SPDC applies to a wide range of modern SPDC experiments with entangled photons without the need for additional approximations. For both the narrowband and broadband regimes, analytical formulas and compact closed-form expressions for the biphoton state can be derived, as shown in detail in [29, 56]. The main idea of the derivation is based on expressing the

The experimental parameters used to tune the expansion amplitudes can be grouped into three categories: the transverse spatial distribution of the pump beam, the structure of the nonlinear crystal, and the geometry of the measurement setup. When inserting Vp(qs+qi)=LGℓppp(qs+qi,w)V_p(\boldsymbol{q}_s+\boldsymbol{q}_i)=\mathrm{LG}_{\ell_p}^{p_p}(\boldsymbol{q}_s+\boldsymbol{q}_i,w), we obtain (in the narrowband regime)

Laguerre polynomials as a sum,

LGpℓ(q,w)=e−∣q∣2w24eiℓArg⁡(q)∑u=0pTup,ℓ(w)∣q∣2u+∣ℓ∣\mathrm{LG}_{p}^{\ell}(\boldsymbol{q},w)=e^{-\frac{|\boldsymbol{q}|^{2}w^{2}}{4}}e^{i\ell\operatorname{Arg}(\boldsymbol{q})}\sum_{u=0}^{p}T_{u}^{p,\ell}(w)|\boldsymbol{q}|^{2u+|\ell|}

with the polynomial expansion coefficient

Tup,ℓ(w)=p!(p+∣ℓ∣)!π(w2)2u+∣ℓ∣+1(−1)p+uℓp(p−u)!(∣ℓ∣+u)!u!,T_{u}^{p,\ell}(w)=\sqrt{\frac{p!(p+|\ell|)!}{\pi}}\left(\frac{w}{\sqrt{2}}\right)^{2u+|\ell|+1}\frac{(-1)^{p+u}\ell^{p}}{(p-u)!(|\ell|+u)!u!},

which allows deriving analytical expressions in polar coordinates. If the nonlinear crystal is instead pumped by an arbitrary transverse field expressed by LG modes,

Vp=∑pp,ℓpapp,ℓpLG⁡(pp,ℓp,wp),V_p=\sum_{p_p,\ell_p}a_{p_p,\ell_p}\operatorname{LG}(p_p,\ell_p,w_p),

the amplitudes are updated according to the linearity of the SPDC amplitude in the pump field to

Cps,piℓs,ℓi(app,ℓp,wp)∝∑pp,ℓpapp,ℓpCps,piℓs,ℓi(pp,ℓp,wp).C_{p_s,p_i}^{\ell_s,\ell_i}(a_{p_p,\ell_p},w_p)\propto\sum_{p_p,\ell_p}a_{p_p,\ell_p}C_{p_s,p_i}^{\ell_s,\ell_i}(p_p,\ell_p,w_p).

The coefficients are normalized according to ∑pp,ℓp∣app,ℓp∣2=1\sum_{p_p,\ell_p}|a_{p_p,\ell_p}|^2=1. Although the pp dependence was analyzed in great detail, there is no corresponding conservation rule. For experimental convenience, the pp index is often omitted, and therefore it has often been referred to as the ”forgotten quantum number” [51]. In the single Schmidt mode regime (SSMR), the correlations of higher-order radial modes p>0p>0 of the OAM spectrum nearly vanish for pp=0p_p=0, and the dominant contribution to the state is concentrated in the mode ps=pi=0p_s=p_i=0. Figure 5 shows the same distribution as Fig. 1(b) and (c) of the main text, illustrating this behavior. The SSMR is achieved when the Rayleigh range of the pump beam is equal to half the crystal length and the ratio between the pump and collection waists is set to γ=wpws=wpwi=12\gamma=\frac{w_p}{w_s}=\frac{w_p}{w_i}=\frac{1}{\sqrt{2}}, as pointed out in [38, 59]. By optimizing the down-conversion regime for a minimal spiral bandwidth, we can maximize the global fidelity and justify our choice of target modes with p=0p=0.

OAM spectra compared for different pump beam waists

Figure 5. OAM spectra compared for different pump beam waists. OAM spectra pumped with a superposition of OAM indices ℓp=0,3\ell_p=0,3 like in Fig. 1 of the main text, but radial indices ps=pi=0,1p_s=p_i=0,1 are included. In the SSMR, most mode distribution is concentrated for the lowest radial order, minimizing the extent of the spiral bandwidth. The regime can be achieved for a waist ratio γ=wpws,i=12\gamma=\frac{w_p}{w_{s,i}}=\frac{1}{\sqrt{2}}. Here, we set ws,i=25w_{s,i}=25 µm.

global fidelity and justify our choice of target modes with p=0p=0.

Inserting the Kronecker delta into Eq. (4) and setting the radial indices of the signal and idler photons to be zero, such that ∣0,ℓs⟩:=∣ℓs⟩|0,\ell_s\rangle:=|\ell_s\rangle and ∣ℓi⟩:=∣0,ℓi⟩=∣0,ℓp−ℓs⟩|\ell_i\rangle:=|0,\ell_i\rangle=|0,\ell_p-\ell_s\rangle leads to expression Eq. (1) from the main text if C0,0ℓs,ℓi≡C0,0ℓs,ℓp−ℓs:=CℓsℓiC_{0,0}^{\ell_s,\ell_i}\equiv C_{0,0}^{\ell_s,\ell_p-\ell_s}:=C_{\ell_s}^{\ell_i} is defined. Additionally, by writing a0,ℓp:=aℓpa_{0,\ell_p}:=a_{\ell_p} we recover Eq. (3) from the main text. To calculate the global fidelity, we use the full state, including higher radial orders, to accurately quantify how much of the target state population matches the experimentally expected spectra.

Relative pump weights

The advantage of our proposed method is that it determines the relative pump weights using only the coincidence amplitudes Cps,piℓs,ℓi(pp,ℓp,wp)C_{p_s,p_i}^{\ell_s,\ell_i}(p_p,\ell_p,w_p). Here, we provide additional details on how to calculate the pump weights aℓpa_{\ell_p} and explain why including higher radial pump modes is disadvantageous. As an example, we reconstruct the MES shown in Fig. 1 of the main text. The target MES in OAM 13(∣0,0⟩+∣1,2⟩+∣2,1⟩)\frac{1}{\sqrt{3}}(|0,0\rangle+|1,2\rangle+|2,1\rangle) (for radial indices ps=pi=0p_s=p_i=0) within the continuous subspace S3=span⁡{∣0⟩,∣1⟩,∣2⟩}\mathcal{S}_3=\operatorname{span}\{|0\rangle,|1\rangle,|2\rangle\} can be written as a matrix of coefficients. By assuming the same value of wpw_p for all pump modes, we obtain a coefficient matrix

(C0,00,0C0,00,1C0,00,2C0,01,0C0,01,1C0,01,2C0,02,0C0,02,1C0,02,2)⟶(a0,0C0,00,0(0,0)a0,1C0,00,1(0,1)a0,2C0,00,2(0,2)a0,1C0,01,0(0,1)a0,2C0,01,1(0,2)a0,3C0,01,2(0,3)a0,2C0,02,0(0,2)a0,3C0,02,1(0,3)a0,4C0,02,2(0,4))=13(100001010).\begin{pmatrix} C_{0,0}^{0,0} & C_{0,0}^{0,1} & C_{0,0}^{0,2}\\ C_{0,0}^{1,0} & C_{0,0}^{1,1} & C_{0,0}^{1,2}\\ C_{0,0}^{2,0} & C_{0,0}^{2,1} & C_{0,0}^{2,2} \end{pmatrix} \longrightarrow \begin{pmatrix} a_{0,0}C_{0,0}^{0,0}(0,0) & a_{0,1}C_{0,0}^{0,1}(0,1) & a_{0,2}C_{0,0}^{0,2}(0,2)\\ a_{0,1}C_{0,0}^{1,0}(0,1) & a_{0,2}C_{0,0}^{1,1}(0,2) & a_{0,3}C_{0,0}^{1,2}(0,3)\\ a_{0,2}C_{0,0}^{2,0}(0,2) & a_{0,3}C_{0,0}^{2,1}(0,3) & a_{0,4}C_{0,0}^{2,2}(0,4) \end{pmatrix} =\frac{1}{\sqrt{3}} \begin{pmatrix} 1&0&0\\ 0&0&1\\ 0&1&0 \end{pmatrix}.

After the arrow, we introduced a general pump superposition of OAM modes (with pp=0p_p=0) relevant in the subspace: Vp(pp,ℓp)=a0,0 LG(0,0)+a0,1 LG(0,1)+a0,2 LG(0,2)+a0,3 LG(0,3)+a0,4 LG(0,4)V_p(p_p,\ell_p)=a_{0,0}\,\mathrm{LG}(0,0)+a_{0,1}\,\mathrm{LG}(0,1)+a_{0,2}\,\mathrm{LG}(0,2)+a_{0,3}\,\mathrm{LG}(0,3)+a_{0,4}\,\mathrm{LG}(0,4). Here, we assume that the projected state is normalized within the target subspace, so that the target amplitudes are 1/31/\sqrt{3}.

The choice a0,1=a0,2=a0,4=0a_{0,1}=a_{0,2}=a_{0,4}=0 is immediate. For the other pump weights, we find a0,0=(3C0,00,0(0,0))−1a_{0,0}=\left(\sqrt{3}C^{0,0}_{0,0}(0,0)\right)^{-1} as well as a0,3=(3C0,02,1(0,3))−1a_{0,3}=\left(\sqrt{3}C^{2,1}_{0,0}(0,3)\right)^{-1}. The degeneracy condition kp/2=ks=kik_p/2=k_s=k_i along with ws=wiw_s=w_i ensures C0,01,2(0,3)=C0,02,1(0,3)C^{1,2}_{0,0}(0,3)=C^{2,1}_{0,0}(0,3), leading to the desired state. The corresponding OAM spectrum is again shown in Figure 6 (a). Since we are in the SSMR, the contributions are predominantly concentrated within the ps=pi=0p_s=p_i=0 sector.

We might expect that the same procedure can be extended to a pump superposition including modes with radial index pp=1p_p=1. It is then important to emphasize which pump mode gives rise to which two-photon amplitude. Although we obtain a similar system of equations,

(a1,0C0,00,0(1,0)a1,1C0,00,1(1,1)a1,2C0,00,2(1,2)a1,1C0,01,0(1,1)a1,2C0,01,1(1,2)a1,3C0,01,2(1,3)a1,2C0,02,0(1,2)a1,3C0,02,1(1,3)a1,4C0,02,2(1,4))=!13(100001010)\begin{pmatrix} a_{1,0}C^{0,0}_{0,0}(1,0) & a_{1,1}C^{0,1}_{0,0}(1,1) & a_{1,2}C^{0,2}_{0,0}(1,2) \\ a_{1,1}C^{1,0}_{0,0}(1,1) & a_{1,2}C^{1,1}_{0,0}(1,2) & a_{1,3}C^{1,2}_{0,0}(1,3) \\ a_{1,2}C^{2,0}_{0,0}(1,2) & a_{1,3}C^{2,1}_{0,0}(1,3) & a_{1,4}C^{2,2}_{0,0}(1,4) \end{pmatrix} \mathrel{\overset{!}{=}} \frac{1}{\sqrt{3}} \begin{pmatrix} 1 & 0 & 0 \\ 0 & 0 & 1 \\ 0 & 1 & 0 \end{pmatrix}

and find equivalent solutions a1,1=a1,2=a1,4=0a_{1,1}=a_{1,2}=a_{1,4}=0, a1,0=(3C0,00,0(1,0))−1a_{1,0}=\left(\sqrt{3}C^{0,0}_{0,0}(1,0)\right)^{-1}, a1,3=(3C0,02,1(1,3))−1a_{1,3}=\left(\sqrt{3}C^{2,1}_{0,0}(1,3)\right)^{-1}, we observe that the fidelity is practically zero, as shown in Figure 6 (b). The reason is that we are optimizing only within a restricted subspace. Within the target subspace, the requirement of equalized amplitudes is fulfilled, as apparent from the inset. Globally, since C0,00,0(1,0)≪C0,0−1,1(1,0)C^{0,0}_{0,0}(1,0)\ll C^{-1,1}_{0,0}(1,0), the resulting weights lead to 13C0,0−1,1(1,0)C0,00,0(1,0)≫13\frac{1}{\sqrt{3}}\frac{C^{-1,1}_{0,0}(1,0)}{C^{0,0}_{0,0}(1,0)}\gg\frac{1}{\sqrt{3}}. Consequently, the above choice results in an inefficient state preparation. As discussed in the main text, it is therefore preferable to restrict to those modes with maximal amplitudes generated by the pump and to use them as target modes.

For a pump LG(1,0)\mathrm{LG}(1,0) the maximum amplitudes are C0,10,0=C1,00,0C^{0,0}_{0,1}=C^{0,0}_{1,0}. Interestingly, the solution leads mainly to the state 12(∣0,−1;0,1⟩+∣0,1;0,−1⟩+∣1,0;0,0⟩+∣0,0;1,0⟩)\frac{1}{2}(|0,-1;0,1\rangle+|0,1;0,-1\rangle+|1,0;0,0\rangle+|0,0;1,0\rangle), using the abbreviation ∣ps,ℓs⟩⊗∣pi,ℓi⟩:=∣ps,ℓs;pi,ℓi⟩|p_s,\ell_s\rangle\otimes|p_i,\ell_i\rangle:=|p_s,\ell_s;p_i,\ell_i\rangle. However, this state is not maximally entangled within the subspace span⁡{∣p,ℓ⟩}\operatorname{span}\{|p,\ell\rangle\}, p=0,1p=0,1, ℓ=−1,0,1\ell=-1,0,1. See also Sec. F for further details.

Similarly, a pump superposition with different radial indices, e.g., with pp=0,1p_p=0,1, leads to a system of equations of the form

(a0,0C0,00,0(0,0)+a1,0C0,00,0(1,0)a0,1C0,00,1(0,1)+a1,1C0,00,1(1,1)a0,2C0,00,2(0,2)+a1,2C0,00,2(1,2)a0,1C0,01,0(0,1)+a1,1C0,01,0(1,1)a0,2C0,01,1(0,2)+a1,2C0,01,1(1,2)a0,3C0,01,2(0,3)+a1,3C0,01,2(1,3)a0,2C0,02,0(0,2)+a1,2C0,02,0(1,2)a0,3C0,02,1(0,3)+a1,3C0,02,1(1,3)a0,4C0,02,2(0,4)+a1,4C0,02,2(1,4))=!13(100001010)\begin{pmatrix} a_{0,0}C^{0,0}_{0,0}(0,0)+a_{1,0}C^{0,0}_{0,0}(1,0) & a_{0,1}C^{0,1}_{0,0}(0,1)+a_{1,1}C^{0,1}_{0,0}(1,1) & a_{0,2}C^{0,2}_{0,0}(0,2)+a_{1,2}C^{0,2}_{0,0}(1,2) \\ a_{0,1}C^{1,0}_{0,0}(0,1)+a_{1,1}C^{1,0}_{0,0}(1,1) & a_{0,2}C^{1,1}_{0,0}(0,2)+a_{1,2}C^{1,1}_{0,0}(1,2) & a_{0,3}C^{1,2}_{0,0}(0,3)+a_{1,3}C^{1,2}_{0,0}(1,3) \\ a_{0,2}C^{2,0}_{0,0}(0,2)+a_{1,2}C^{2,0}_{0,0}(1,2) & a_{0,3}C^{2,1}_{0,0}(0,3)+a_{1,3}C^{2,1}_{0,0}(1,3) & a_{0,4}C^{2,2}_{0,0}(0,4)+a_{1,4}C^{2,2}_{0,0}(1,4) \end{pmatrix} \mathrel{\overset{!}{=}} \frac{1}{\sqrt{3}} \begin{pmatrix} 1 & 0 & 0 \\ 0 & 0 & 1 \\ 0 & 1 & 0 \end{pmatrix}

In this case, we can set again a0,1=a1,1=a0,2=a1,2=a0,4=a1,4=0a_{0,1}=a_{1,1}=a_{0,2}=a_{1,2}=a_{0,4}=a_{1,4}=0, but there are infinitely many solutions for the remaining pump weights. For example, choosing a0,0=0.95⋅(3C0,00,0(0,0))−1a_{0,0}=0.95\cdot\left(\sqrt{3}C^{0,0}_{0,0}(0,0)\right)^{-1}, a1,0=0.05⋅(3C0,00,0(1,0))−1a_{1,0}=0.05\cdot\left(\sqrt{3}C^{0,0}_{0,0}(1,0)\right)^{-1}, a0,3=0.95⋅(3C0,02,1(0,3))−1a_{0,3}=0.95\cdot\left(\sqrt{3}C^{2,1}_{0,0}(0,3)\right)^{-1}, and a1,3=0.05⋅(3C0,02,1(1,3))−1a_{1,3}=0.05\cdot\left(\sqrt{3}C^{2,1}_{0,0}(1,3)\right)^{-1} yields the result shown in Fig. 7 (a). For the same reason as before, we conclude that the solution a1,0=a1,3=0a_{1,0}=a_{1,3}=0 is the most efficient choice.

This also clarifies the dimensionality limit of MES in continuous subspaces discussed in the main text. For the subspace S7=span⁡{∣−3⟩,…,∣3⟩}\mathcal{S}_7=\operatorname{span}\{|{-3}\rangle,\ldots,|3\rangle\}, we showed in Figure 2 (g) of the main text an OAM spectrum, where the center modes along the anti-diagonal ℓp=0\ell_p=0 need to be amplitude-equalized, C0,0−1,1=C0,00,0=C0,01,−1C^{-1,1}_{0,0}=C^{0,0}_{0,0}=C^{1,-1}_{0,0}. Fig. 7 (b) shows the OAM spectrum of an MES 17(∣−3,−2⟩+∣−2,−3⟩+∣−1,1⟩+∣0,0⟩+∣1,−1⟩+∣2,3⟩+∣3,2⟩)\frac{1}{\sqrt{7}}(|-3,-2\rangle+|-2,-3\rangle+|-1,1\rangle+|0,0\rangle+|1,-1\rangle+|2,3\rangle+|3,2\rangle) with Schmidt number K=6.91\mathcal{K}=6.91 obtained for a pump superposition additionally including the terms a1,0 LG1,0a_{1,0}\,\mathrm{LG}_{1,0}, a2,0 LG2,0a_{2,0}\,\mathrm{LG}_{2,0}, a3,0 LG3,0a_{3,0}\,\mathrm{LG}_{3,0} to equalize the three modes along the anti-diagonal ℓp=0\ell_p=0. When viewed globally, the fidelity is reduced since the global maxima are the modes with amplitudes C1,00,0=C0,10,0C^{0,0}_{1,0}=C^{0,0}_{0,1}. Increasing the dimensionality dd further leads to increasingly complex scenarios and systems of equations. Inverse design algorithms, such as those shown in [42, 43], provide a powerful alternative in such cases, but come at the cost of losing the analytical simplicity. Our approach, based on disassembled subspaces, is an efficient alternative.

OAM spectra pumped with a superposition of mixed OAM and radial indices

Figure 7. OAM spectra pumped with a superposition of mixed OAM and radial indices. (a) OAM spectrum for a spatial pump distribution with OAM modes ℓp=0,3\ell_p = 0, 3 and pp=0,1p_p = 0, 1. Similar to Figure 6, the maxima may lie outside of the intended subspace, resulting in a small global fidelity. (b) The same pump distribution as shown in Figure 2(g) of the main text, with additional higher radial modes pp=1,2,3p_p = 1, 2, 3 for ℓp=0\ell_p = 0 to overcome the continuous subspace limit. Now, within S7×7\mathcal{S}_{7 \times7} with S7=span⁡{∣0⟩,∣±1⟩,∣±2⟩,∣±3⟩}\mathcal{S}_7 = \operatorname{span}\{|0\rangle, |\pm1\rangle, |\pm2\rangle, |\pm3\rangle\}, we can find a nearly MES, K=6.91\mathcal{K} = 6.91 and Fglob=0.39\mathcal{F}_{\mathrm{glob}} = 0.39. The resulting maxima are again outside of the target subspace and reduce the global fidelity.

When calculating pump weights for a desired superposition in the broadband regime, the spectral dependence of the coincidence amplitudes must be taken into account, since Cps,piℓs,ℓi(pp,ℓp,wp)C^{\ell_s,\ell_i}_{p_s,p_i}(p_p,\ell_p,w_p) depends on Ωs\Omega_s and Ωi\Omega_i [48, 64]. To generate a state analogous to that in Figure 6(a), with equalized modes across different anti-diagonals in the broadband regime, the spectral dependence of the coincidence amplitudes must be made explicit, leading to

a0,0=(3∫dΩsdΩi∣C0,00,0(0,0,wp,Ωs,Ωi)∣2)−1/2a_{0,0} = \left(3\int d\Omega_s d\Omega_i \left|C^{0,0}_{0,0}(0,0,w_p,\Omega_s,\Omega_i)\right|^2\right)^{-1/2}

and

a0,3=(3∫dΩsdΩi∣C0,01,2(0,3,wp,Ωs,Ωi)∣2)−1/2a_{0,3} = \left(3\int d\Omega_s d\Omega_i \left|C^{1,2}_{0,0}(0,3,w_p,\Omega_s,\Omega_i)\right|^2\right)^{-1/2}

where the spectral filter bandwidth restricts the integration over the measured frequency band.

In the broadband regime, these weights equalize the frequency-integrated populations of the selected OAM modes. This condition alone does not automatically guarantee a pure spatial MES: after tracing over frequency, the coherence between target terms is determined by the overlap of their joint spectral amplitudes. Nevertheless, Ref. [56] showed that all mode pairs sharing the same relative mode number, NR=2pp−2ps−2pi+∣ℓp∣−∣ℓs∣−∣ℓi∣N_R = 2p_p - 2p_s - 2p_i + |\ell_p| - |\ell_s| - |\ell_i|, possess identical spectral dependence, such that the spatial and spectral degrees of freedom are separable. Since all target modes considered and suggested in our work satisfy NR=0N_R = 0, the resulting maximally entangled states remain spatially pure.

Comparison with large pump waists

If the pump beam waist wpw_p is chosen much larger than the projection waists of the signal and idler ws=wiw_s = w_i, the spiral bandwidth can be significantly broadened, leading to a more uniformly distributed OAM spectrum. In this section, we briefly compare our approach of generating MESs with broadening the OAM spectrum through a large pump waist.

The waist configuration is commonly characterized by the ratio γ=wp/ws=wp/wi\gamma= w_p/w_s = w_p/w_i. For the SSMR, this ratio needs to be close to γ=12\gamma= \frac{1}{\sqrt{2}}. As an illustrative example, we consider a Gaussian pump (ℓp=0\ell_p = 0) with a ratio γ=20\gamma= 20, as shown in Figure 8(a). Compared with all previous examples, the resulting spiral bandwidth is substantially broader and remains significant even for higher radial indices ps=pip_s = p_i. As before, we can identify subspaces in which the modes are uniformly distributed. Since all modes are confined to a single anti-diagonal, the conditions for an MES are automatically satisfied if the amplitudes are equalized. We select two examples: one MES restricted to the zeroth radial mode and a second MES including higher radial indices.

OAM spectra for large pump beam waist

Figure 8. OAM spectra for large pump beam waist. (a) The pump beam carries zero OAM ℓp=0\ell_p = 0. For a large waist ratio γ=wp/ws=wp/wi=20\gamma= w_p/w_s = w_p/w_i = 20, the spiral bandwidth is substantially broadened and more uniformly distributed. Non-negligible amplitudes are even maintained for high radial indices psp_s and pip_i. (b) If we select the subspace S11=span⁡{∣−5⟩,∣−4⟩,…,∣5⟩}S_{11} = \operatorname{span}\{\lvert-5\rangle, \lvert-4\rangle, \ldots, \lvert5\rangle\} with ps=pi=0p_s = p_i = 0, we can obtain a 11×1111 \times11 MES. However, globally, the large number of undesired modes outside the subspace renders this an inefficient method for preparing MESs. (c) We can alternatively include radial modes ps=pi=1p_s = p_i = 1 within the subspace of OAM modes span⁡{∣−2⟩,…,∣2⟩}\operatorname{span}\{\lvert-2\rangle, \ldots, \lvert2\rangle\} since the spiral bandwidth is so broad, the central modes are matched too. Maximal entanglement within the selected subspace is achieved. Nevertheless, the global fidelity remains low due to the significant population outside the target subspace.

In Figure 8(b), we consider the subspace S11=span⁡{∣−5⟩,∣−4⟩,…,∣4⟩,∣5⟩}\mathcal{S}_{11} = \operatorname{span}\{|{-5}\rangle, |{-4}\rangle, \ldots, |4\rangle, |5\rangle\} and the state

∣ψ⟩=111(∣−5,5⟩+∣−4,4⟩+⋯+∣4,−4⟩−∣5,−5⟩).|\psi\rangle= \frac{1}{\sqrt{11}}\left(|{-5},5\rangle+ |{-4},4\rangle+ \cdots+ |4,{-4}\rangle- |5,{-5}\rangle\right).

Within the subspace, the modes are equally distributed, as quantified by a Schmidt number of K=10.99\mathcal{K} = 10.99 and a fidelity of Fsub=1.00\mathcal{F}_{\mathrm{sub}} = 1.00. Because the broadened spiral bandwidth also equalizes the amplitudes for modes with ps=pi=1p_s = p_i = 1, we can enlarge with respect to higher pp modes. The state @page 11/15 starts_mid=1 ends_mid=1

OAM spectra pumped with a superposition of mixed OAM and radial indices

Figure 7 (PDF p. 11). OAM spectra pumped with a superposition of mixed OAM and radial indices. (a) OAM spectrum for a spatial pump distribution with OAM modes ℓp=0,3\ell_p = 0, 3 and pp=0,1p_p = 0, 1. Similar to Figure 6, the maxima may lie outside of the intended subspace, resulting in a small global fidelity. (b) The same pump distribution as shown in Figure 2(g) of the main text, with additional higher radial modes pp=1,2,3p_p = 1, 2, 3 for ℓp=0\ell_p = 0 to overcome the continuous subspace limit. Now, within S7×7\mathcal{S}_{7 \times7} with S7=span⁡{∣0⟩,∣±1⟩,∣±2⟩,∣±3⟩}\mathcal{S}_7 = \operatorname{span}\{|0\rangle, |\pm1\rangle, |\pm2\rangle, |\pm3\rangle\}, we can find a nearly MES, K=6.91\mathcal{K} = 6.91 and Fglob=0.39\mathcal{F}_{\mathrm{glob}} = 0.39. The resulting maxima are again outside of the target subspace and reduce the global fidelity.

efficient alternative.

When calculating pump weights for a desired superposition in the broadband regime, the spectral dependence of the coincidence amplitudes must be taken into account, since Cps,piℓs,ℓi(pp,ℓp,wp)C^{\ell_s,\ell_i}_{p_s,p_i}(p_p,\ell_p,w_p) depends on Ωs\Omega_s and Ωi\Omega_i [48, 64]. To generate a state analogous to that in Figure 6(a), with equalized modes across different anti-diagonals in the broadband regime, the spectral dependence of the coincidence amplitudes must be made explicit, leading to

a0,0=(3∫dΩsdΩi∣C0,00,0(0,0,wp,Ωs,Ωi)∣2)−1/2a_{0,0} = \left(3\int d\Omega_s d\Omega_i \left|C^{0,0}_{0,0}(0,0,w_p,\Omega_s,\Omega_i)\right|^2\right)^{-1/2}

and

a0,3=(3∫dΩsdΩi∣C0,01,2(0,3,wp,Ωs,Ωi)∣2)−1/2a_{0,3} = \left(3\int d\Omega_s d\Omega_i \left|C^{1,2}_{0,0}(0,3,w_p,\Omega_s,\Omega_i)\right|^2\right)^{-1/2}

where the spectral filter bandwidth restricts the integration over the measured frequency band.

In the broadband regime, these weights equalize the frequency-integrated populations of the selected OAM modes. This condition alone does not automatically guarantee a pure spatial MES: after tracing over frequency, the coherence between target terms is determined by the overlap of their joint spectral amplitudes. Nevertheless, Ref. [56] showed that all mode pairs sharing the same relative mode number, NR=2pp−2ps−2pi+∣ℓp∣−∣ℓs∣−∣ℓi∣N_R = 2p_p - 2p_s - 2p_i + |\ell_p| - |\ell_s| - |\ell_i|, possess identical spectral dependence, such that the spatial and spectral degrees of freedom are separable. Since all target modes considered and suggested in our work satisfy NR=0N_R = 0, the resulting maximally entangled states remain spatially pure.

Comparison with large pump waists

If the pump beam waist wpw_p is chosen much larger than the projection waists of the signal and idler ws=wiw_s = w_i, the spiral bandwidth can be significantly broadened, leading to a more uniformly distributed OAM spectrum. In this section, we briefly compare our approach of generating MESs with broadening the OAM spectrum through a large pump waist.

The waist configuration is commonly characterized by the ratio γ=wp/ws=wp/wi\gamma= w_p/w_s = w_p/w_i. For the SSMR, this ratio needs to be close to γ=12\gamma= \frac{1}{\sqrt{2}}. As an illustrative example, we consider a Gaussian pump (ℓp=0\ell_p = 0) with a ratio γ=20\gamma= 20, as shown in Figure 8(a). Compared with all previous examples, the resulting spiral bandwidth is substantially broader and remains significant even for higher radial indices ps=pip_s = p_i. As before, we can identify subspaces in which the modes are uniformly distributed. Since all modes are confined to a single anti-diagonal, the conditions for an MES are automatically satisfied if the amplitudes are equalized. We select two examples: one MES restricted to the zeroth radial mode and a second MES including higher radial indices.

In Figure 8(b), we consider the subspace S11=span⁡{∣−5⟩,∣−4⟩,…,∣4⟩,∣5⟩}\mathcal{S}_{11} = \operatorname{span}\{|{-5}\rangle, |{-4}\rangle, \ldots, |4\rangle, |5\rangle\} and the state

∣ψ⟩=111(∣−5,5⟩+∣−4,4⟩+⋯+∣4,−4⟩−∣5,−5⟩).|\psi\rangle= \frac{1}{\sqrt{11}}\left(|{-5},5\rangle+ |{-4},4\rangle+ \cdots+ |4,{-4}\rangle- |5,{-5}\rangle\right).

Within the subspace, the modes are equally distributed, as quantified by a Schmidt number of K=10.99\mathcal{K} = 10.99 and a fidelity of Fsub=1.00\mathcal{F}_{\mathrm{sub}} = 1.00. Because the broadened spiral bandwidth also equalizes the amplitudes for modes with ps=pi=1p_s = p_i = 1, we can enlarge with respect to higher pp modes. The state

∣ψ⟩=110(∣0,−2;0,2⟩−∣0,−1;0,1⟩+⋯+∣0,2;0,−2⟩+∣1,−2;1,2⟩−∣1,−1;1,1⟩+⋯+∣1,2;1,−2⟩)\lvert\psi\rangle= \frac{1}{\sqrt{10}}\left(\lvert0,-2;0,2\rangle- \lvert0,-1;0,1\rangle+ \cdots+ \lvert0,2;0,-2\rangle+ \lvert1,-2;1,2\rangle- \lvert1,-1;1,1\rangle+ \cdots+ \lvert1,2;1,-2\rangle\right)

is maximally entangled in the subspace S10=span⁡{∣0,−2⟩,∣0,−1⟩,∣0,0⟩,∣0,1⟩,∣0,2⟩,∣1,−2⟩,∣1,−1⟩,∣1,0⟩,∣1,1⟩,∣1,2⟩}S_{10} = \operatorname{span}\{\lvert0,-2\rangle, \lvert0,-1\rangle, \lvert0,0\rangle, \lvert0,1\rangle, \lvert0,2\rangle, \lvert1,-2\rangle, \lvert1,-1\rangle, \lvert1,0\rangle, \lvert1,1\rangle, \lvert1,2\rangle\} as evident from K=9.99\mathcal{K} = 9.99 and Fsub=1.00\mathcal{F}_{\mathrm{sub}} = 1.00.

However, the global fidelities reveal that exploiting a broad spiral bandwidth is ultimately inefficient. While a larger pump waist increases the number of nearly equalized modes, it also distributes a substantial amount of the photon-pair modes into the long tails of the spectrum outside the chosen subspace. Consequently, the global fidelities are much smaller with only Fglob=0.011\mathcal{F}_{\mathrm{glob}} = 0.011 and Fglob=0.019\mathcal{F}_{\mathrm{glob}} = 0.019 for the considered two states in Figure 8, respectively. We note that in a broadband regime, the state is not spatially pure as the target modes have different RMNs.

Detailed description of Figure 4 (b)

Figure 4 (b) in the main text summarizes the findings for MESs up to dimensionality d=12d = 12. Tab. I lists the pump and subspace configurations selected using the design criteria of Sec. III B, together with their subspace fidelities, Schmidt numbers, and global fidelities. In particular, we specify which pump superpositions and subspace modes are recommended.

Include higher radial indices of signal and idler

Recent advances in optical mode detection techniques, such as LG mode sorters [52], have enabled the resolution of both pp and ℓ\ell mode indices, for full modal characterization of LG states. Experimental demonstrations have reported efficient sorting for radial and OAM modes using various strategies, such as those based on the Gouy phase [53], or fractional Fourier transforms [54, 55]. Therefore, it is also of interest to explore the possibility of generating MES in subspaces that include both ℓ\ell and higher pp modes.

The established design approach based on disassembled subspaces can be extended to subspaces that cover higher radial modes. Instead of enlarging the disassembled subspace through successively higher pump OAM values, one may alternatively introduce higher radial orders into the LG superposition. However, since no conservation law exists for pp modes, the generated distribution of radial signal and idler modes is less intuitive. Consequently, selecting suitable modes to form the subspace becomes less obvious since the dominant amplitudes cannot be identified a priori. Nevertheless, we observe a systematic trend that provides a useful guideline: in the SSMR for even ppp_p, the dominant amplitude occurs in the mode sector with ps=pi=pp2p_s = p_i = \frac{p_p}{2}. We now again use the notation ∣pj,ℓj⟩|p_j,\ell_j\rangle for a single photon j=s,ij = s, i and ∣ps,ℓs;pi,ℓi⟩|p_s,\ell_s;p_i,\ell_i\rangle for two photons.

ddVp=∑aℓpLG0,ℓp,ℓp=V_p = \sum a_{\ell_p}\mathrm{LG}_{0,\ell_p}, \ell_p =Sd=span⁡{∣ℓ⟩},ℓ=S_d = \operatorname{span}\{|\ell\rangle\}, \ell=Fsub\mathcal{F}_{\mathrm{sub}}K\mathcal{K}Fglob\mathcal{F}_{\mathrm{glob}}
30, 30, 1, 21.003.000.679
4±3\pm3±1,±2\pm1, \pm21.004.000.623
50, ±5\pm50, ±2,±3\pm2, \pm30.9984.980.560
6±3,7\pm3, 7±1,±2,3,4\pm1, \pm2, 3, 40.9965.960.556
7±3,7,−6\pm3, 7, -6±1,±2,3,4,−3\pm1, \pm2, 3, 4, -30.9976.960.478
8±3,±7\pm3, \pm7±1,±2,±3,±4\pm1, \pm2, \pm3, \pm40.9957.910.525
9±3,±7,14\pm3, \pm7, 14±1,±2,±3,±4,7\pm1, \pm2, \pm3, \pm4, 70.9958.910.452
10±3,±7,15\pm3, \pm7, 15±1,±2,±3,±4,7,8\pm1, \pm2, \pm3, \pm4, 7, 80.9949.900.473
11±3,±7,−14,15\pm3, \pm7, -14, 15±1,±2,±3,±4,±7,8\pm1, \pm2, \pm3, \pm4, \pm7, 80.99510.890.411
12±3,±7,±15\pm3, \pm7, \pm15±1,±2,±3,±4,±7,±8\pm1, \pm2, \pm3, \pm4, \pm7, \pm80.99511.890.370

Table 1. Data corresponding to Figure 4 (b).

We look at an example for d=8d = 8 and start with the same OAM indices as in Figure 4 of the main text for the single-photon subspace S4=span⁡{∣0,±1⟩,∣0,±2⟩}S_4 = \operatorname{span}\{|0,\pm1\rangle, |0,\pm2\rangle\}. The modes generated by OAM pump modes a0,±3a_{0,\pm3} are still located in the upper-left sector of the OAM spectrum. Allowing ps=pi=0,1p_s = p_i = 0,1 effectively doubles the dimensionality. The total subspace S8=span⁡{∣0,±2⟩,∣0,±1⟩,∣1,±1⟩,∣1,±2⟩}S_8 = \operatorname{span}\{|0,\pm2\rangle, |0,\pm1\rangle, |1,\pm1\rangle, |1,\pm2\rangle\} is highlighted in Figure 9 in red. By adding suitable weights a2,±3a_{2,\pm3}, we generate the most probable modes C1,11,1=C−1,−21,1C_{1,1}^{1,1} = C_{-1,-2}^{1,1} and C1,11,−2=C1,1−2,−1C_{1,1}^{1,-2} = C_{1,1}^{-2,-1} in the subspace part with ps=pi=1p_s = p_i = 1.

OAM spectra of a $d = 8$ maximally entangled state in a subspace with radial indices

Figure 9. OAM spectra of a d=8d = 8 MES in a subspace with radial indices for signal and idler ps=pi=0,1p_s = p_i = 0, 1. A pump superposition containing both OAM and radial modes enables a subspace-optimized MES whose target modes extend into sectors with nonzero signal and idler radial indices.

We have to check that undesired, non-vanishing amplitudes within the sector ps=0,pi=1p_s = 0,p_i = 1 and ps=1,pi=0p_s = 1,p_i = 0 do not fall within the constructed two-photon Hilbert space, as this would reduce the Schmidt number and subspace fidelity. Employing radial pump modes with pp≠0p_p \ne0 increases the number of non-negligible contributions of signal and idler modes with higher radial indices. Compared to the states shown in Figure 4 (b) of the main text, this reduces the global fidelity. This trend becomes more pronounced as dd is further increased and, therefore, less attractive for an efficient experimental preparation of MESs.

References

References

  1. [1]S. Slussarenko and G. J. Pryde, Photonic quantum information processing: A concise review, Appl. Phys. Rev. 6 (2019).DOI
  2. [2]F. Flamini, N. Spagnolo, and F. Sciarrino, Photonic quantum information processing: a review, Rep. Prog. Phys. 82, 016001 (2019).DOI
  3. [3]H.-S. Zhong, H. Wang, Y.-H. Deng, M.-C. Chen, L.-C. Peng, Y.-H. Luo, J. Qin, D. Wu, X. Ding, Y. Hu, et al., Quantum computational advantage using photons, Science 370, 1460 (2020).DOI
  4. [4]M. Erhard, M. Krenn, and A. Zeilinger, Advances in high-dimensional quantum entanglement, Nat. Rev. Phys. 2, 365 (2020).DOI
  5. [5]D. Cozzolino, B. Da Lio, D. Bacco, and L. K. Oxenløwe, High-dimensional quantum communication: benefits, progress, and future challenges, Adv. Quantum Technol. 2, 1900038 (2019).DOI
  6. [6]F. Zhu, M. Tyler, N. H. Valencia, M. Malik, and J. Leach, Is high-dimensional photonic entanglement robust to noise?, AVS Quantum Sci. 3 (2021). (2012).
  7. [7]Y. Wang, Z. Hu, B. C. Sanders, and S. Kais, Qudits and high-dimensional quantum computing, Front. Phys. (Lausanne) 8, 589504 (2020).DOI
  8. [8]V. D’Ambrosio, N. Spagnolo, L. Del Re, S. Slussarenko, Y. Li, L. C. Kwek, L. Marrucci, S. P. Walborn, L. Aolita, and F. Sciarrino, Photonic polarization gears for ultra-sensitive angular measurements, Nat. Commun. 4, 2432 (2013).DOI
  9. [9]A. K. Jha, G. S. Agarwal, and R. W. Boyd, Supersensitive measurement of angular displacements using entangled photons, Phys. Rev. A 83, 053829 (2011).DOI
  10. [10]L. Allen, M. W. Beijersbergen, R. J. C. Spreeuw, and J. P. Woerdman, Orbital angular momentum of light and the transformation of laguerre-gaussian laser modes, Phys. Rev. A 45, 8185 (1992).
  11. [11]G. Molina-Terriza, J. P. Torres, and L. Torner, Twisted photons, Nat. Phys. 3, 305 (2007).DOI
  12. [12]A. M. Yao and M. J. Padgett, Orbital angular momentum: origins, behavior and applications, Adv. Opt. Photonics 3, 161 (2011).DOI
  13. [13]M. Erhard, R. Fickler, M. Krenn, and A. Zeilinger, Twisted photons: new quantum perspectives in high dimensions, Light Sci. Appl. 7, 17146 (2018).DOI
  14. [14]A. Forbes and I. Nape, Quantum mechanics with patterns of light: progress in high dimensional and multidimensional entanglement with structured light, AVS Quantum Sci. 1 (2019).DOI
  15. [15]M. Krenn, R. Fickler, M. Fink, J. Handsteiner, M. Malik, T. Scheidl, R. Ursin, and A. Zeilinger, Communication with spatially modulated light through turbulent air across vienna, New J. Phys. 16, 113028 (2014).
  16. [16]M. Mirhosseini, O. S. Magaña-Loaiza, M. N. O’Sullivan, B. Rodenburg, M. Malik, M. P. Lavery, M. J. Padgett, D. J. Gauthier, and R. W. Boyd, High-dimensional quantum cryptography with twisted light, New J. Phys. 17, 033033 (2015).
  17. [17]M. Krenn, J. Handsteiner, M. Fink, R. Fickler, and A. Zeilinger, Twisted photon entanglement through turbulent air across vienna, Proc. Natl. Acad. Sci. U.S.A. 112, 14197 (2015).DOI
  18. [18]J. Wang, J.-Y. Yang, I. M. Fazal, N. Ahmed, Y. Yan, H. Huang, Y. Ren, Y. Yue, S. Dolinar, M. Tur, et al., Terabit free-space data transmission employing orbital angular momentum multiplexing, Nat. Photonics 6, 488 (2012).DOI
  19. [19]N. Bozinovic, Y. Yue, Y. Ren, M. Tur, P. Kristensen, H. Huang, A. E. Willner, and S. Ramachandran, Terabit-scale orbital angular momentum mode division multiplexing in fibers, Science 340, 1545 (2013).DOI
  20. [20]P.-A. Moreau, E. Toninelli, T. Gregory, and M. J. Padgett, Imaging with quantum states of light, Nat. Rev. Phys. 1, 367 (2019).DOI
  21. [21]A. Anwar, C. Perumangatt, F. Steinlechner, T. Jennewein, and A. Ling, Entangled photon-pair sources based on three-wave mixing in bulk crystals, Rev. Sci. Instrum. 92 (2021).DOI
  22. [22]A. Mair, A. Vaziri, G. Weihs, and A. Zeilinger, Entanglement of the orbital angular momentum states of photons, Nature 412, 313 (2001).pubmed.ncbi.nlm.nih.gov/11460157
  23. [23]R. Fickler, R. Lapkiewicz, W. N. Plick, M. Krenn, C. Schaeff, S. Ramelow, and A. Zeilinger, Quantum entanglement of high angular momenta, Science 338, 640 (2012).DOI
  24. [24]H. Di Lorenzo Pires, H. C. B. Florijn, and M. P. van Exter, Measurement of the spiral spectrum of entangled two-photon states, Phys. Rev. Lett. 104, 020505 (2010).
  25. [25]M. Krenn, M. Huber, R. Fickler, R. Lapkiewicz, S. Ramelow, and A. Zeilinger, Generation and confirmation of a (100×100)-dimensional entangled quantum system, Proc. Natl. Acad. Sci. U.S.A. 111, 6243 (2014).DOI
  26. [26]R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, Quantum entanglement, Rev. Mod. Phys. 81, 865 (2009).DOI
  27. [27]C. Xu, S. Huang, Q. Yu, D. Wei, P. Chen, S. Nie, Y. Zhang, and M. Xiao, Manipulating the orbital-angular-momentum correlation of entangled two-photon states in three-dimensional nonlinear photonic crystals, Phys. Rev. A 104, 063716 (2021).DOI
  28. [28]E. Rozenberg, A. Karnieli, O. Yesharim, J. Foley-Comer, S. Trajtenberg-Mills, D. Freedman, A. M. Bronstein, and A. Arie, Inverse design of spontaneous parametric down-conversion for generation of high-dimensional qudits, Optica 9, 602 (2022).DOI
  29. [29]R. Bernecker, B. Baghdasaryan, and S. Fritzsche, High-dimensional maximally entangled photon pairs in parametric down-conversion, Phys. Rev. A 110, 033718 (2024).DOI
  30. [30]A. Vaziri, J.-W. Pan, T. Jennenwein, G. Weihs, and A. Zeilinger, Concentration of higher dimensional entanglement: qutrits of photon orbital angular momentum, Phys. Rev. Lett. 91, 227902 (2003).pubmed.ncbi.nlm.nih.gov/14683274
  31. [31]A. C. Dada, J. Leach, G. S. Buller, M. J. Padgett, and E. Andersson, Experimental high-dimensional two-photon entanglement and violations of generalized bell inequalities, Nat. Phys. 7, 677 (2011).DOI
  32. [32]F. Wang, M. Erhard, A. Babazadeh, M. Malik, M. Krenn, and A. Zeilinger, Generation of the complete four-dimensional bell basis, Optica 4, 1462 (2017).DOI
  33. [33]Y. Cheng, W. Zhang, D. X. Qiu, and L. Chen, Coherent generation of the complete high-dimensional bell basis by adaptive pump modulation, Phys. Rev. Appl. 14, 054069 (2020).
  34. [34]S. Liu, Y. Zhang, C. Yang, S. Liu, Z. Ge, Y. Li, Y. Li, Z. Zhou, G. Guo, and B. Shi, Increasing two-photon entangled dimensions by shaping input-beam profiles, Phys. Rev. A 101, 052324 (2020).DOI
  35. [35]P. Boucher, H. Defienne, and S. Gigan, Engineering spatial correlations of entangled photon pairs by pump beam shaping, Opt. Lett. 46, 4200 (2021).DOI
  36. [36]J. P. Torres, Y. Deyanova, L. Torner, and G. Molina-Terriza, Preparation of engineered two-photon entangled states for multidimensional quantum information, Phys. Rev. A 67, 052313 (2003).DOI
  37. [37]A. M. Yao, Angular momentum decomposition of entangled photons with an arbitrary pump, New J. Phys. 13, 053048 (2011).DOI
  38. [38]E. V. Kovlakov, S. S. Straupe, and S. P. Kulik, Quantum state engineering with twisted photons via adaptive shaping of the pump beam, Phys. Rev. A 98, 060301(R) (2018).DOI
  39. [39]S. Liu, Z. Zhou, S. Liu, Y. Li, Y. Li, C. Yang, Z. Xu, Z. Liu, G. Guo, and B. Shi, Coherent manipulation of a three-dimensional maximally entangled state, Phys. Rev. A 98, 062316 (2018).
  40. [40]N. Bornman, W. Tavares Buono, M. Lovemore, and A. Forbes, Optimal pump shaping for entanglement con C. Xu, L. Xiao, J. He, H. Zhang, D. Wei, X. Hu, M. Xiao, and Y. Zhang, Manipulating the radial components of lg pump beam for ultrahigh-dimensional maximally entangled orbital angular momentum states, Opt. Express 30, 11120 (2022).
  41. [42]S. Karan, R. Prasad, and A. K. Jha, Postselection-free controlled generation of a high-dimensional orbital-angular-momentum entangled state, Phys. Rev. Appl. 20, 054027 (2023).DOI
  42. [43]J. F. Comer, O. Yesharim, S. Mishra, S. Prabhakar, E. Rozenberg, A. Karnieli, R. P. Singh, and A. Arie, Inverse-design-based experimental generation of high-dimensional spatially entangled light, Optica Quantum 4, 179 (2026).DOI
  43. [44]L. Kopf, R. Barros, S. Prabhakar, E. Giese, and R. Fickler, Conservation of angular momentum on a single-photon level, Phys. Rev. Lett. 134, 203601 (2025).DOI
  44. [45]S. Walborn, C. Monken, S. Pádua, and P. Souto Ribeiro, Spatial correlations in parametric down-conversion, Phys. Rep. 495, 87 (2010).DOI
  45. [46]F. M. Miatto, A. M. Yao, and S. M. Barnett, Full characterization of the quantum spiral bandwidth of entangled biphotons, Phys. Rev. A 83, 033816 (2011).DOI
  46. [47]S. Karan, S. Aarav, H. Bharadwaj, L. Taneja, A. De, G. Kulkarni, N. Meher, and A. K. Jha, Phase matching in β-barium borate crystals for spontaneous parametric down-conversion, Journal of Optics 22, 083501 (2020).DOI
  47. [48]R. Bernecker, B. Baghdasaryan, and S. Fritzsche, Spatial and temporal characteristics of spontaneous parametric down-conversion with varying focal planes of interacting beams, Eur. Phys. J. D 77, 172 (2023).DOI
  48. [49]J. P. Torres, A. Alexandrescu, S. Carrasco, and L. Torner, Quasi-phase-matching engineering for spatial control of entangled two-photon states, Opt. Lett. 29, 376 (2004).
  49. [50]A. Dosseva, L. Cincio, and A. M. Brańczyk, Shaping the joint spectrum of down-converted photons through optimized custom poling, Phys. Rev. A 93, 018301 (2016).DOI
  50. [51]W. N. Plick and M. Krenn, Physical meaning of the radial index of laguerre-gauss beams, Phys. Rev. A 92, 063841 (2015).
  51. [52]N. K. Fontaine, R. Ryf, H. Chen, D. T. Neilson, K. Kim, and J. Carpenter, Laguerre-gaussian mode sorter, Nat. Commun. 10, 1865 (2019).
  52. [53]phase radial mode sorter for light: concepts and experiments, Phys. Rev. Lett. 120, 103601 (2018).DOI
  53. [54]Y. Zhou, M. Mirhosseini, D. Fu, J. Zhao, S. M. Hashemi Rafsanjani, A. E. Willner, and R. W. Boyd, Sorting photons by radial quantum number, Phys. Rev. Lett. 119, 263602 (2017).DOI
  54. [55]D. Fu, Y. Zhou, R. Qi, S. Oliver, Y. Wang, S. M. H. Rafsanjani, J. Zhao, M. Mirhosseini, Z. Shi, P. Zhang, et al., Realization of a scalable laguerre–gaussian mode sorter based on a robust radial mode sorter, Opt. Express 26, 33057 (2018).DOI
  55. [56]B. Baghdasaryan, C. Sevilla-Gutiérrez, F. Steinlechner, and S. Fritzsche, Generalized description of the spatio-temporal biphoton state in spontaneous parametric down-conversion, Phys. Rev. A 106, 063711 (2022).DOI
  56. [57]L. Zhu and J. Wang, Arbitrary manipulation of spatial amplitude and phase using phase-only spatial light modulators, Sci. Rep. 4, 7441 (2014).DOI
  57. [58]J. P. Torres, A. Alexandrescu, and L. Torner, Quantum spiral bandwidth of entangled two-photon states, Phys. Rev. A 68, 050301(R) (2003).DOI
  58. [59]E. Kovlakov, I. Bobrov, S. Straupe, and S. Kulik, Spatial bell-state generation without transverse mode subspace postselection, Phys. Rev. Lett. 118, 030503 (2017).DOI
  59. [60]R. Bernecker, B. Baghdasaryan, and S. Fritzsche, Engineering of maximally entangled orbital angular momentum states via path identity, Phys. Rev. A 112, 063701 (2025).DOI
  60. [61]F. M. Miatto, H. Di LorenzoPires, S. M. Barnett, and M. P. van Exter, Spatial Schmidt modes generated in parametric down-conversion, Eur. Phys. J. D 66, 1 (2012).DOI
  61. [62]F. M. Miatto, T. Brougham, and A. M. Yao, Cartesian and polar schmidt bases for down-converted photons: How high dimensional entanglement protects the shared information from non-ideal measurements, Eur. Phys. J. D 66, 1 (2012).
  62. [63]B. Baghdasaryan and S. Fritzsche, Enhanced entanglement from ince-gaussian pump beams in spontaneous parametric down-conversion, Phys. Rev. A 102, 052412 (2020).DOI
  63. [64]C. Sevilla-Gutiérrez, V. R. Kaipalaith, B. Baghdasaryan, M. Gräfe, S. Fritzsche, and F. Steinlechner, Spectral properties of transverse laguerre-gauss modes in parametric down-conversion, Phys. Rev. A 109, 023534 (2024).

Paper details

vlnc:W_5p5sbdhk
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 4

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents