Skip to main content
Log in

Abstract

Quantum cloning is a fundamental protocol of quantum information theory. Perfect universal quantum cloning is prohibited by the laws of quantum mechanics, only imperfect copies being reachable. Symmetric quantum cloning is concerned with case when the quality of the clones is identical. In this work, we study the general case of \(1 \rightarrow N\) asymmetric cloning, where one asks for arbitrary qualities of the clones. We characterize, for all Hilbert space dimensions and number of clones, the set of all possible clone qualities. This set is realized as the nonnegative part of the unit ball of a newly introduced norm, which we call the \(\mathcal Q\)-norm. We also provide a closed-form expression for the quantum cloner achieving a given clone quality vector. Our analysis relies on the Schur–Weyl duality and on the study of the spectral properties of partially transposed permutation operators.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Data Availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.

Notes

  1. In \(\mathcal {V}^\lambda \), the vectors \(\left| \textbf{v}^{\lambda }_j\right\rangle \) are defined on different tensor products, depending on the value of i in their corresponding \(\left| \Omega \right\rangle _{(0,i)}\).

References

  1. Bennett, C.H., Brassard, G.: Quantum cryptography: Public key distribution and coin tossing. arXiv preprint arXiv:2003.06557 (2020)

  2. Bridgeman, J.C., Chubb, C.T.: Hand-waving and interpretive dance: an introductory course on tensor networks. J. Phys. A: Math. Theor. 50(22), 223001 (2017)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  3. Bruß, D., Cinchetti, M., D’Ariano, G.M., Macchiavello, C.: Phase-covariant quantum cloning. Phys. Rev. A 62(1), 012302 (2000)

    Article  ADS  Google Scholar 

  4. Barnum, H., Caves, C.M., Fuchs, C.A., Jozsa, R., Schumacher, B.: Noncommuting mixed states cannot be broadcast. Phys. Rev. Lett. 76(15), 2818 (1996)

    Article  ADS  Google Scholar 

  5. Bruß, D., DiVincenzo, D.P., Ekert, A., Fuchs, C.A., Macchiavello, C., Smolin, J.A.: Optimal universal and state-dependent quantum cloning. Phys. Rev. A 57(4), 2368 (1998)

    Article  ADS  Google Scholar 

  6. Bruss, D., D’Ariano, G.M., Macchiavello, C., Sacchi, M.F.: Approximate quantum cloning and the impossibility of superluminal information transfer. Phys. Rev. A 62(6), 062302 (2000)

    Article  ADS  Google Scholar 

  7. Bužek, V., Hillery, M.: Quantum copying: Beyond the no-cloning theorem. Phys. Rev. A 54(3), 1844 (1996)

    Article  ADS  MathSciNet  Google Scholar 

  8. BuŽek, V., Hillery, M.: Universal optimal cloning of qubits and quantum registers. In: NASA International Conference on Quantum Computing and Quantum Communications, pp. 235–246. Springer, (1998)

  9. Bhatia, R., Kittaneh, F.: Norm inequalities for partitioned operators and an application. Math. Ann. 287, 719–726 (1990)

    Article  MathSciNet  MATH  Google Scholar 

  10. Coyle, B., Doosti, M., Kashefi, E., Kumar, N.: Variational quantum cloning: improving practicality for quantum cryptanalysis. arXiv preprint arXiv:2012.11424 (2020)

  11. Cerf, N.J.: Asymmetric quantum cloning in any dimension. J. Mod. Opt. 47(2–3), 187–209 (2000)

    Article  ADS  MathSciNet  Google Scholar 

  12. Ćwikliński, P., Horodecki, M., Studziński, M.: Region of fidelities for a 1 \(\rightarrow \) n universal qubit quantum cloner. Phys. Lett. A 376(32), 2178–2187 (2012)

    Article  ADS  MathSciNet  Google Scholar 

  13. Coecke, B., Kissinger, A.: Picturing quantum processes. Cambridge University Press, Cambridge (2017)

    Book  MATH  Google Scholar 

  14. Coecke, B., Kissinger, A.: Picturing quantum processes. In: International Conference on Theory and Application of Diagrams, pp. 28–31. Springer (2018)

  15. Christandl, M., Leditzky, F., Majenz, C., Smith, G., Speelman, F., Walter, M.: Asymptotic performance of port-based teleportation. Commun. Math. Phys. 381, 379–451 (2021)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  16. Du, J., Durt, T., Zou, P., Li, H., Kwek, L.C., Lai, C.H., Oh, C.H., Ekert, A.: Experimental quantum cloning with prior partial information. Phys. Rev. Lett. 94(4), 040505 (2005)

    Article  ADS  Google Scholar 

  17. Duan, L.-M., Guo, G.-C.: Probabilistic cloning and identification of linearly independent quantum states. Phys. Rev. Lett. 80(22), 4999 (1998)

    Article  ADS  Google Scholar 

  18. Dieks, D.G.B.J.: Communication by epr devices. Phys. Lett. A 92(6), 271–272 (1982)

    Article  ADS  Google Scholar 

  19. Duncan, R., Kissinger, A., Perdrix, S., Van De Wetering, J.: Graph-theoretic simplification of quantum circuits with the zx-calculus. Quantum 4, 279 (2020)

    Article  Google Scholar 

  20. Fiurasek, J., Filip, R., Cerf, N.J.: Highly asymmetric quantum cloning in arbitrary dimension. arXiv preprint arXiv:quant-ph/0505212 (2005)

  21. Fulton, W., Harris, J.: Representation Theory: A First Course, vol. 129. Springer, Berlin (2013)

    MATH  Google Scholar 

  22. Fan, H., Wang, Y.-N., Jing, L., Yue, J.-D., Shi, H.-D., Zhang, Y.-L., Liang-Zhu, M.: Quantum cloning machines and the applications. Phys. Rep. 544(3), 241–322 (2014)

    Article  ADS  MathSciNet  Google Scholar 

  23. Gisin, N., Massar, S.: Optimal quantum cloning machines. Phys. Rev. Lett. 79(11), 2153 (1997)

    Article  ADS  Google Scholar 

  24. Grinko, D., Ozols, M.: Linear programming with unitary-equivariant constraints. arXiv preprint arXiv:2207.05713 (2022)

  25. Hashagen, A.K.: Universal Asymmetric Quantum Cloning Revisited. arXiv preprint arXiv:1607.03723 (2016)

  26. Kay, A.: Optimal Universal Quantum Cloning: Asymmetries and Fidelity Measures. arXiv preprint arXiv:1407.4951 (2014)

  27. Kay, A., Kaszlikowski, D., Ramanathan, R.: Optimal cloning and singlet monogamy. Phys. Rev. Lett. 103(5), 050501 (2009)

    Article  ADS  Google Scholar 

  28. Kay, A., Ramanathan, R., Kaszlikowski, D.: Optimal asymmetric quantum cloning. arXiv preprint arXiv:1208.5574 (2012)

  29. Keyl, M., Werner, R.F.: Optimal cloning of pure states, testing single clones. J. Math. Phys. 40(7), 3283–3299 (1999)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  30. Leditzky, F.: Optimality of the pretty good measurement for port-based teleportation. Lett. Math. Phys. 112(5), 98 (2022)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  31. Lostaglio, M., Senno, G.: Contextual advantage for state-dependent cloning. Quantum 4, 258 (2020)

    Article  Google Scholar 

  32. Mozrzymas, M., Horodecki, M., Studziński, M.: Structure and properties of the algebra of partially transposed permutation operators. J. Math. Phys. 55(3), 032202 (2014)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  33. Mozrzymas, M., Studziński, M., Horodecki, M.: A simplified formalism of the algebra of partially transposed permutation operators with applications. J. Phys. A: Math. Theor. 51(12), 125202 (2018)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  34. Mozrzymas, M., Studziński, M., Strelchuk, S., Horodecki, M.: Optimal port-based teleportation. New J. Phys. 20(5), 053006 (2018)

    Article  ADS  Google Scholar 

  35. Nielsen, M.A., Chuang, I.L.: Quantum Computation and Quantum Information. Cambridge University Press, Cambridge (2010)

    MATH  Google Scholar 

  36. Nechita, I., Pellegrini, C., Rochette, D.: A geometrical description of the universal \(1 \rightarrow 2\) asymmetric quantum cloning region. Quantum Inf. Process. 20(10), 1–18 (2021)

    Article  MathSciNet  MATH  Google Scholar 

  37. Penrose, R.: Applications of negative dimensional tensors. Comb. Math. Its Appl. 1, 221–244 (1971)

    MathSciNet  MATH  Google Scholar 

  38. Studziński, M., Ćwikliński, P., Horodecki, M., Mozrzymas, M.: Group-representation approach to 1\(\rightarrow \) n universal quantum cloning machines. Phys. Rev. A 89(5), 052322 (2014)

    Article  ADS  Google Scholar 

  39. Studziński, M., Horodecki, M., Mozrzymas, M.: Commutant structure of \(u^{\otimes (n - 1)} \otimes u^*\) transformations. J. Phys. A: Math. Theor. 46(39), 395303 (2013)

    Article  MathSciNet  MATH  Google Scholar 

  40. Scarani, V., Iblisdir, S., Gisin, N., Acin, A.: Quantum cloning. Rev. Mod. Phys. 77(4), 1225 (2005)

    Article  ADS  MathSciNet  MATH  Google Scholar 

  41. Studziński, M., Strelchuk, S., Mozrzymas, M., Horodecki, M.: Port-based teleportation in arbitrary dimension. Sci. Rep. 7(1), 10871 (2017)

    Article  ADS  Google Scholar 

  42. Watrous, J.: The Theory of Quantum Information. Cambridge University Press, Cambridge (2018)

    Book  MATH  Google Scholar 

  43. Wood, C.J., Biamonte, J.D., Cory, D.G.: Tensor networks and graphical calculus for open quantum systems. arXiv preprint arXiv:1111.6950 (2011)

  44. Wood, C.J., Biamonte, J.D., Cory, D.G.: Tensor networks and graphical calculus for open quantum systems. Quantum Inf. Comput. 15(9–10), 759–811 (2015)

    MathSciNet  Google Scholar 

  45. Werner, R.F.: Optimal cloning of pure states. Phys. Rev. A 58(3), 1827 (1998)

    Article  ADS  Google Scholar 

  46. Wootters, W.K., Zurek, W.H.: A single quantum cannot be cloned. Nature 299(5886), 802–803 (1982)

    Article  ADS  MATH  Google Scholar 

Download references

Acknowledgements

We thank the reviewers for the careful reading of our manuscript and for their useful and constructive remarks on our work. The authors were supported by the ANR project ESQuisses, grant number ANR-20-CE47-0014-01 as well as by the PHC programs Sakura (Random Matrices and Tensors for Quantum Information and Machine Learning) and Procope (Entanglement Preservation in Quantum Information Theory). I.N. has also received support from the ANR project STARS, grant number ANR-20-CE40-0008. C.P is also supported by the ANR projects Q-COAST ANR- 19-CE48-0003, “Quantum Trajectories” ANR- 20-CE40-0024-01, and “Investissements d’Avenir” ANR-11-LABX-0040 of the French National Research Agency.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Ion Nechita.

Ethics declarations

Conflict of interest

On behalf of all authors, the corresponding author states that there is no conflict of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendices

Appendix A: Structure of permutation operators

This appendix is devoted to the presentation of technical results on permutations operator. Numerous results or remarks are illustrated by graphical calculus.

1.1 graphical calculus

To illustrate the technical results of this section, we employ a graphical method for representing tensors, which is based on Penrose’s graphical notation [37]. Comparable approaches have been established in the context of tensor network states and categorical quantum information theory, as detailed in [2, 13, 44].

In this diagrammatic notation, tensors are depicted as boxes and wires.

figure a

The wires are labeled with indices, and a box represents the tensor’s value corresponding to those specific indices.

figure b

Specifically, a vector is represented by a box with a single wire extending to the left, labeled by the coordinate index. Conversely, a dual vector has its wire directed to the right.

figure c

Note that the left and right directions follow the convention that corresponds to the standard right-to-left composition in linear algebra.

Tensor diagrams can be combined using two methods: tensor product and tensor contraction. The tensor product combines two diagrams either vertically or horizontally.

figure d

Tensor contraction combines two diagrams by taking the sum across all common indices.

figure e

Specifically, we recover the scalar product.

figure f

The matrix product.

figure g

And the matrix trace.

figure h

Scalars multiply the diagrams and are displayed adjacent to the diagram.

figure i

In conclusion, the transposition can be visually illustrated by exchanging the “input” and “output” wires of a box representing a matrix. The partial transposition involves only the permutation of wires corresponding to specific spaces.

figure j

1.2 Contributions of permutations

In Sect. 3, we only focus on permutations where 0 is not a fixed point. This is motivated by the following remarks

Let \(T: \mathcal {M}_d \rightarrow {(\mathcal {M}_d)}^{\otimes N}\) be a \(\mathcal {U}(d)\)-covariant quantum channel, and \(C_T\) its Choi matrix, then from Proposition 3.2 we have for all \(\rho \in \mathcal {D}^1_d\)

$$\begin{aligned} T(\rho )&= \hbox {Tr}_0 \Big [ C_T \big ( \rho ^\textsf{T}\otimes I^{\otimes N} \big ) \Big ] \\&= \hbox {Tr}_0 \bigg [ \sum _{\sigma \in \mathfrak {S}_{N + 1}} \beta _\sigma \cdot \Pi _\sigma ^{\mathsf {\Gamma }} \big ( \rho ^\textsf{T}\otimes I^{\otimes N} \big ) \bigg ], \end{aligned}$$

for some \(\beta _\sigma \in \mathbb {C}\) (see Example A.3 for the case \(N=2\)).

Let us justify that \(\sigma \in \mathfrak {S}_{N + 1}\) such that \(\sigma (0) = 0\) do not contribute to the performance of the cloning map, whereas the \((N + 1)\) cycles contribute the most to the copies. Indeed, for a permutation \(\sigma \in \mathfrak {S}_{N + 1}\) such that \(\sigma (0) = 0\) we have \(\Pi _\sigma ^{\mathsf {\Gamma }} = \Pi _\sigma \) and for all \(\rho \in \mathcal {D}^1_d\)

$$\begin{aligned} \hbox {Tr}_1 \Big [ \Pi _\sigma ^{\mathsf {\Gamma }} \big ( \rho ^\textsf{T}\otimes I^{\otimes N} \big ) \Big ]&= \hbox {Tr}_0 \Big [ \Pi _\sigma \big ( \rho ^\textsf{T}\otimes I^{\otimes N} \big ) \Big ] \\&= \frac{\hbox {Tr}\rho }{d} \hbox {Tr}_0 \Big [ \Pi _\sigma \; I^{\otimes (N + 1)} \Big ] \\&= \frac{1}{d} \cdot \hbox {Tr}_0 \Pi _\sigma \end{aligned}$$

whose contributions consist in deteriorating the quality of the cloning, since the quantity \(\hbox {Tr}_1 \sigma \) does not depend on \(\rho \). Now, if \(\sigma \in \mathfrak {S}_{N + 1}\) is a \((N + 1)\) cycle, the marginals of

$$\begin{aligned} T_{\scriptscriptstyle \text {perfect}} : \mathcal {M}_d&\longrightarrow {(\mathcal {M}_d)}^{\otimes N} \\ \rho&\longmapsto \hbox {Tr}_0 \Big [ \Pi _\sigma ^{\mathsf {\Gamma }} \big ( \rho ^\textsf{T}\otimes I^{\otimes N} \big ) \Big ] \end{aligned}$$

are all equals to identity, i.e., \({(T_{\scriptscriptstyle \text {perfect}})}_i(\rho ) = \rho \) for all \(\rho \in \mathcal {M}_d\). This would be the case of a perfect quantum cloning map, but the operator \(T_{\scriptscriptstyle \text {perfect}}\) are of course not completely positive operator due to the no-cloning theorem (see Example A.4 for computations with graphical calculus).

More generally, let \(T: \mathcal {M}_d \rightarrow {(\mathcal {M}_d)}^{\otimes N}\) be a \(\mathcal {U}(d)\)-covariant quantum channel. From Proposition 3.2, the marginals \(T_i\) are for all \(\rho \in \mathcal {D}^1_d\)

$$\begin{aligned} T_i(\rho ) = p_i \! \cdot \! \rho + (1 - p_i)\frac{I}{d} \end{aligned}$$

where the \(p_i\) are the sums of the \(\frac{N!}{2}\) coefficients \(\beta _\sigma \) of the permutations contributing for this copy, that is, the \(\sigma \) such that i is in the orbit of 0 under the action of \(\sigma \). The following examples illustrate some results in terms of graphical calculus.

Example A.1

(Partition of \(\Sigma _{a,b}, N = 3\)) The permutation set \(\big \{ \Pi ^{\mathsf {\Gamma }}_\sigma \; \big | \;\sigma \in \mathfrak {S}_4, \quad \sigma (0) \ne 0 \big \}\) is partitioned in

Example A.2

(Partially transposed permutations, \(N = 3\)) On \(\mathfrak {S}_{4}\), the \((0 \, i)\)-partially transposed permutations are

Example A.3

(Choi Matrix \(N = 2\)) Let \(T: \mathcal {M}_d \rightarrow \mathcal {M}_d \otimes \mathcal {M}_d\) be a \(\mathcal {U}(d)\)-covariant quantum channel, then its Choi matrix \(C_T\) is:

Example A.4

(Contributions of quantum cloning) Let \(T: \mathcal {M}_d \rightarrow \mathcal {M}_d \otimes \mathcal {M}_d\) be a \(\mathcal {U}(d)\)-covariant quantum channel, such that its Choi matrix \(C_T\) is the partially transposed cycle \((1 \, 2 \, 3)\):

Then, the two marginals are

and

1.3 Proof of Lemma lem:struct

In this subsection, we prove the key Lemma 3.4, which gives the structure of partially permutations operators and action on vectors. We illustrate as well with graphical calculus.

Lemma A.5

Let \(1 \le a,b \le N\), for any \(\sigma \in \Sigma _{a,b}\), there exists a unique \(\hat{\sigma } \in \mathfrak {S}_{N - 1}\) such that the partially transposed permutation operator \(\sigma ^{\mathsf {\Gamma }}\) is

$$\begin{aligned} \Pi _\sigma ^{\mathsf {\Gamma }} = \Pi _{(1 \, a)} \, (\omega _{(0,1)} \otimes \Pi _{\hat{\sigma }}) \; \Pi _{(1 \, b)}. \end{aligned}$$

Proof

Let \(1 \le a,b,c,d \le N\), and \(\sigma \in \Sigma _{a,b}\), then \(\Pi _{(c \, a)} \; \Pi _\sigma ^{\mathsf {\Gamma }} \; \Pi _{(d \, b)}\) is a partially transposed permutation operator from \(\Sigma _{c,d}\). Let \(\sigma ' \in \Sigma _{1,1}\) be the permutation such that

$$\begin{aligned} \Pi _{(1 \, a)} \; \Pi _\sigma ^{\mathsf {\Gamma }} \; \Pi _{(1 \, b)} = \Pi _{\sigma '}^{\mathsf {\Gamma }}. \end{aligned}$$

There is a unique \(\hat{\sigma } \in \mathfrak {S}_{N - 1}\), such that

$$\begin{aligned} \Pi _{\sigma '}. = \Pi _{(0,1)} \otimes \Pi _{\hat{\sigma }}, \end{aligned}$$

and finally

$$\begin{aligned} \Pi _{\sigma '}^{\mathsf {\Gamma }}. = \omega _{(0,1)} \otimes \Pi _{\hat{\sigma }}, \end{aligned}$$

where the permutation operator \(\Pi _{\hat{\sigma }}\) is

$$\begin{aligned} \Pi _{\hat{\sigma }} = \frac{1}{d} \cdot \hbox {Tr}_{0,1} \big [ \Pi _{(1 \, a)} \; \Pi _\sigma ^{\mathsf {\Gamma }} \; \Pi _{(1 \, b)} \big ]. \end{aligned}$$

\(\square \)

Example A.6

Let \(\sigma = (0 \, 3 \, 2) \in \mathfrak {S}_4\). Its partial transposition operator

can be decomposed into

Recall that for two integers ij, we denote by \((i \,: \, j)\) the permutation \(\mathfrak {S}_n\) defined by

$$\begin{aligned} (i \, : \, j) = {\left\{ \begin{array}{ll} \big ( (i - 1) \, (i - 2) \, \cdots \, j \big ) &{}\text {if } 0 \le j< i \le n \\ \big ( i \, (i + 1) \, \cdots \, (j - 1) \big ) &{}\text {if } 0 \le i < j \le n \\ (0)(1) \cdots (n - 1) &{}\text {otherwise} \end{array}\right. }. \end{aligned}$$

That is, the \(|i - j|\) cycle of \(\mathfrak {S}_n\) between i and j.

Example A.7

The non-trivial \((i \,: \, j)\)-cycles of \(\mathfrak {S}_3\) are:

Lemma A.8

Let \(1 \le a,b,c \le N\), and let \(\sigma \in \Sigma _{a,b}\). Let \(\phi \in \mathcal {H}^{\otimes (N - 1)}\), then

$$\begin{aligned} \Pi _\sigma ^{\mathsf {\Gamma }}&(\left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle ) \\&= {\left\{ \begin{array}{ll} d \cdot \left| \Omega \right\rangle _{(0,a)} \otimes \big ( \Pi _{(0 \, : \, (a - 1))} \circ \Pi _{\hat{\sigma }}\circ \Pi _{((b - 1) \, : \, 0)} \big ) \left| \phi \right\rangle &{}\text {if } b = c \\[1em] \left| \Omega \right\rangle _{(0,a)} \otimes \big ( \Pi _{(0 \, : \, (a - 1))} \circ \Pi _{\hat{\sigma }} \circ \Pi _{((b - 1) \, : \, 0)} \circ \Pi _{((c - 1) \, : \, (b - 1))} \big ) \left| \phi \right\rangle &{}\text {if } b \ne c. \end{array}\right. } \end{aligned}$$

Proof

From Lemma A.5, we can write \(\Pi _\sigma ^{\mathsf {\Gamma }} = \Pi _{(1 \, a)} \, (\omega _{(0,1)} \otimes \Pi _{\hat{\sigma }}) \; \Pi _{(1 \, b)}\), or equivalently

$$\begin{aligned} \Pi _\sigma ^{\mathsf {\Gamma }} = \Pi _{(1 \, a)} \, (I^{\otimes 2} \otimes \Pi _{\hat{\sigma }}) \, \big ( \omega _{(0,1)} \otimes I^{\otimes (N - 1)} \big ) \; \Pi _{(1 \, b)}. \end{aligned}$$

Together with \(\omega \left| \Omega \right\rangle = d \cdot \left| \Omega \right\rangle \), we have the following relations:

$$\begin{aligned} \big ( \omega _{(0,1)} \otimes I^{\otimes (N - 1)} \big ) \; \Pi _{(1 \, b)} = \Pi _{(1 \, b)} \; \big ( \omega _{(0,b)} \otimes I^{\otimes (N - 1)} \big ), \end{aligned}$$

and

$$\begin{aligned} \Pi _{(1 \, b)} \big ( \left| \Omega \right\rangle _{(0,b)} \otimes \left| \phi \right\rangle \big ) = \left| \Omega \right\rangle \otimes \Pi _{((b - 1) \, : \, 0)} \left| \phi \right\rangle . \end{aligned}$$

If \(b = c\), then

$$\begin{aligned}&\big ( \omega _{(0,1)} \otimes I^{\otimes (N - 1)} \big ) \; \Pi _{(1 \, b)} \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big )\\&= \Pi _{(1 \, b)} \; \big ( \omega _{(0,b)} \otimes I^{\otimes (N - 1)} \big ) \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big ) \\&= \Pi _{(1 \, b)} \big ( \omega _{(0,b)} \left| \Omega \right\rangle _{(0,b)} \otimes \left| \phi \right\rangle \big ) \\&= d \cdot \Pi _{(1 \, b)} \big ( \left| \Omega \right\rangle _{(0,b)} \otimes \left| \phi \right\rangle \big ) \\&= d \cdot \big ( \left| \Omega \right\rangle \otimes \Pi _{((b - 1) \, : \, 0)} \left| \phi \right\rangle \big ), \end{aligned}$$

such that finally \(\Pi _\sigma ^{\mathsf {\Gamma }} \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big ) = d \cdot \Pi _{(1 \, a)} \Big [ \left| \Omega \right\rangle _{(0,1)} \otimes \big ( \Pi _{\hat{\sigma }} \circ \Pi _{((b - 1) \,: \, 0)} \circ \big ) \left| \phi \right\rangle \Big ]\). If \(b \ne c\), we have the following relation:

$$\begin{aligned} \big ( \omega _{(0,b)} \otimes I^{\otimes (N - 1)} \big ) \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big ) = \left| \Omega \right\rangle _{(0,b)} \otimes \Pi _\sigma \left| \phi \right\rangle , \end{aligned}$$

for some \(\sigma \in \mathfrak {S}_{N-1}\). Note that \(\left| \phi \right\rangle \) is permuted according to \(\sigma \) which is a cycle whose order depends on whether \(c<b\) or \(b<c\). That is,

$$\begin{aligned} \big ( \omega _{(0,b)} \otimes I^{\otimes (N - 1)} \big ) \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big ) = \left| \Omega \right\rangle _{(0,b)} \otimes \Pi _{((c - 1) \, : \, (b - 1))} \left| \phi \right\rangle . \end{aligned}$$

Then,

$$\begin{aligned} \big ( \omega _{(0,1)} \otimes I^{\otimes (N - 1)} \big ) \; \Pi _{(1 \, b)} \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big )&= \Pi _{(1 \, b)} \; \big ( \omega _{(0,b)} \otimes I^{\otimes (N - 1)} \big ) \big ( \left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle \big ) \\&= \Pi _{(1 \, b)} \big ( \left| \Omega \right\rangle _{(0,b)} \otimes \Pi _{((c - 1) \, : \, (b - 1))} \left| \phi \right\rangle \big ) \\&= \left| \Omega \right\rangle _{(0,1)} \otimes \big ( \Pi _{((b - 1) \, : \, 0)} \circ \Pi _{((c - 1) \, : \, (b - 1))} \big ) \left| \phi \right\rangle , \end{aligned}$$

such that finally 

$$\Pi _\sigma ^{\mathsf {\Gamma }}(\left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle ) = \Pi _{(1 \, a)} \big [ \left| \Omega \right\rangle _{(0,1)} \otimes \big ( \Pi _{\hat{\sigma }} \circ \Pi _{((b - 1) \,: \, 0)} \circ \Pi _{((c - 1) \,: \, (b - 1))} \big ) \left| \phi \right\rangle \Big ].$$

To conclude, from the relation

$$\begin{aligned} \Pi _{(1 \, a)} \big ( \left| \Omega \right\rangle _{(0,1)} \otimes \left| \phi \right\rangle \big ) = \left| \Omega \right\rangle _{(0,a)} \otimes \Pi _{(0 \, : \, (a - 1))} \left| \phi \right\rangle , \end{aligned}$$

we obtain

$$\begin{aligned}&\Pi _\sigma ^{\mathsf {\Gamma }}(\left| \Omega \right\rangle _{(0,c)} \otimes \left| \phi \right\rangle ) \\&= {\left\{ \begin{array}{ll} d \cdot \left| \Omega \right\rangle _{(0,a)} \otimes \big ( \Pi _{(0 \, : \, (a - 1))} \circ \Pi _{\hat{\sigma }}\circ \Pi _{((b - 1) \, : \, 0)} \big ) \left| \phi \right\rangle &{}\text {if } b = c \\[1em] \left| \Omega \right\rangle _{(0,a)} \otimes \big ( \Pi _{(0 \, : \, (a - 1))} \circ \Pi _{\hat{\sigma }} \circ \Pi _{((b - 1) \, : \, 0)} \circ \Pi _{((c - 1) \, : \, (b - 1))} \big ) \left| \phi \right\rangle &{}\text {if } b \ne c. \end{array}\right. } \end{aligned}$$

\(\square \)

Example A.9

Let \(\sigma = (0 \, 3 \, 1 \, 2) \in \mathfrak {S}_4\), and let \(\phi \in \mathcal {H}^{\otimes 2}\), then

In order to simplify the next equations, we use the notations \(\Pi _{\hat{\sigma }(a,b,c)}\) and \(\Pi _{b,c}\), defined by:

$$\begin{aligned} \Pi _{\hat{\sigma }(a,b,c)} = \Pi _{(0 \, : \, (a - 1))} \circ \Pi _{\hat{\sigma }} \circ \Pi _{b,c}, \end{aligned}$$

and

$$\begin{aligned} \Pi _{b,c} = {\left\{ \begin{array}{ll} \Pi _{((b - 1) \, : \, 0)} &{}\text {if } b = c \\ \Pi _{((b - 1) \, : \, 0)} \circ \Pi _{((c - 1) \, : \, (b - 1))} &{}\text {if } b \ne c \end{array}\right. }. \end{aligned}$$

The following lemma concerns the image and the kernel of operators \(\Pi _\sigma ^{\mathsf {\Gamma }}\).

Lemma A.10

For any \(1 \le a,b \le N\) and any \(\sigma \in \Sigma _{a,b}\),

$$\begin{aligned} {{\,\textrm{Im}\,}}\Pi _\sigma ^{\mathsf {\Gamma }}&= {{\,\textrm{Span}\,}}\Big \{ \left| \Omega \right\rangle _{(0,a)} \otimes \left| \phi \right\rangle \; \Big | \; \phi \in \mathcal {H}^{\otimes (N - 1)} \Big \} \\ {{\,\textrm{Ker}\,}}\Pi _\sigma ^{\mathsf {\Gamma }}&\supseteq {{\,\textrm{Span}\,}}{\Big \{ \left| \Omega \right\rangle _{(0,k)} \otimes \left| \phi \right\rangle \; \Big | \; 1 \le k \le N, \; \phi \in \mathcal {H}^{\otimes (N - 1)} \Big \}}^\perp . \end{aligned}$$

Proof

Let \(1 \le a,b \le N\), let \(\sigma \in \Sigma _{a,b}\), and let \(v \in {{\,\textrm{Im}\,}}\Pi _\sigma ^{\mathsf {\Gamma }}\), then \(v = \left| \Omega \right\rangle _{(0,a)} \otimes \left| \phi \right\rangle \) with \(\phi \in \mathcal {H}^{\otimes (N - 1)}\), from Lemma A.5, and

$${{\,\textrm{Im}\,}}\Pi _\sigma ^{\mathsf {\Gamma }} \subseteq {{\,\textrm{Span}\,}}\Big \{ \left| \Omega \right\rangle _{(0,a)} \otimes \left| \phi \right\rangle \; \Big | \; \phi \in \mathcal {H}^{\otimes (N - 1)} \Big \}.$$

Let \(\phi \in \mathcal {H}^{\otimes (N - 1)}_d\), and define . Then from Lemma A.8, \(\Pi _\sigma ^{\mathsf {\Gamma }}(u) = \left| \Omega \right\rangle _{(0,a)} \otimes \left| \phi \right\rangle \), and

$$\begin{aligned} {{\,\textrm{Im}\,}}\Pi _\sigma ^{\mathsf {\Gamma }} \supseteq {{\,\textrm{Span}\,}}\Big \{ \left| \Omega \right\rangle _{(0,a)} \otimes \left| \phi \right\rangle \; \Big | \; \phi \in \mathcal {H}^{\otimes (N - 1)} \Big \}. \end{aligned}$$

Since on finite-dimensional spaces, the kernel of a linear map M is equal to the orthogonal complement of the image of adjoint \(M^*\), i.e., \({{\,\textrm{Ker}\,}}M = {\big ( {{\,\textrm{Im}\,}}M^* \big )}^\perp \), and because the adjoint of \(\Pi _\sigma ^{\mathsf {\Gamma }}\) is a partially transposed permutation operator \(\Pi _{\sigma ^*}^{\mathsf {\Gamma }}\) with \(\sigma ^* \in \Sigma _{b,a}\), we have

$$\begin{aligned} {{\,\textrm{Ker}\,}}\Pi _\sigma ^{\mathsf {\Gamma }}&= {{\,\textrm{Span}\,}}{\Big \{ \left| \Omega \right\rangle _{(1,b)} \otimes \left| \phi \right\rangle \; \Big | \; \phi \in \mathcal {H}^{\otimes (N - 1)} \Big \}}^\perp \\&\supseteq {{\,\textrm{Span}\,}}{\Big \{ \left| \Omega \right\rangle _{(1,k)} \otimes \left| \phi \right\rangle \; \Big | \; 1 \le k \le N,\; \phi \in \mathcal {H}^{\otimes (N - 1)} \Big \}}^\perp \end{aligned}$$

\(\square \)

1.4 Partial trace of partially transposed permutation operators

Lemma A.11

Let \(1 \le a,b \le N\), then

$$\begin{aligned} \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\sigma \in \Sigma _{a,a}} \Pi _\sigma ^{\mathsf {\Gamma }} \bigg ]&= (N - 1)! \; \hbox {Tr}\big [ P^+_{\mathfrak {S}_{N - 1}} \Big ] \cdot I \\ \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma ^{\mathsf {\Gamma }} \bigg ]&= \frac{(N - 1)!}{d} \hbox {Tr}\big [ P^+_{\mathfrak {S}_{N - 1}} \Big ] \cdot I, \end{aligned}$$

where the orthogonal projector onto the trivial representation subspace of \(\mathfrak {S}_{N - 1}\) is defined by \(P^+_{\mathfrak {S}_{N - 1}} = \frac{1}{(N - 1)!} \sum _{\sigma \in \mathfrak {S}_{N - 1}} \Pi _\sigma \).

Proof

For the first equation, we have from Lemma A.5

where \(\Pi _{(0 \, (a - 1))}\) is a permutation operator of \(\mathfrak {S}_N\). For the second equation, since the partial transposition is a linear operator, we have the equality

$$\begin{aligned} \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma ^{\mathsf {\Gamma }} \bigg ] = {\bigg ( \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma \bigg ] \bigg )}^{\mathsf {\Gamma }}. \end{aligned}$$

For any \(\sigma \in \Sigma _{a,b}\), the relation

$$\begin{aligned} (\bar{U} \otimes U^{\otimes N}) \, \Pi _\sigma ^{\mathsf {\Gamma }} \, (\bar{U} \otimes U^{\otimes N}) = \Pi _\sigma ^{\mathsf {\Gamma }} \end{aligned}$$

holds for all \(U \in \mathcal {U}(d)\), such that \(\hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \big [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma ^{\mathsf {\Gamma }} \Big ]\) is a multiple of the identity, and

$$\begin{aligned} \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma ^{\mathsf {\Gamma }} \bigg ] = \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma \bigg ] = c \cdot I, \end{aligned}$$

with \(c = \frac{1}{d} \hbox {Tr}\big [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma \Big ]\). For a permutation \(\sigma \in \mathfrak {S}_n\), we write \(\# \sigma \) the number of disjoint cycles of \(\sigma \). Then,

$$\begin{aligned} \hbox {Tr}\big [ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma \Big ]&= \sum _{\sigma \in \Sigma _{a,b}} d^{\# \sigma } \\&= \sum _{\sigma \in \Sigma _{a,a}} d^{\# [\sigma \circ (a \, b)]}. \end{aligned}$$

Let \(\sigma \in \Sigma _{a,a}\) and write \(\sigma = c_1 \circ \cdots \circ c_k\) the decomposition of \(\sigma \) in disjoint cycles \(c_i = (s^i_1 \, \cdots \, s^i_{l_i})\). If it exists \(i \in [k]\) such that \(a,b \in c_i\), then \(c_i \circ (a \, b)\) can be decomposed in two disjoint cycles. Otherwise, if there exist \(i,j \in [k]\) such that \(a \in c_i\) and \(b \in c_j\), then \(c_i \circ c_j \circ (a \, b)\) can be decomposed in one disjoint cycle. Finally,

$$\begin{aligned} \# \big [ \sigma \circ (a \, b) \Big ] = {\left\{ \begin{array}{ll} \# \sigma + 1 &{}\text {if } \exists i\in [k] \text { s.t. } a,b \in c_i \\ \# \sigma - 1 &{}\text {otherwise} \end{array}\right. }. \end{aligned}$$

But since \(\sigma \in \Sigma _{a,a}\) and \(b \ne a\), in the decomposition of \(\sigma \) in disjoint cycles, a is in the cycle \((0 \, a)\), and \(\# \big [ \sigma \circ (a \, b) \Big ] = \# \sigma - 1\). That is

$$\begin{aligned} \hbox {Tr}\left[ \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma \right] = \frac{1}{d} \hbox {Tr}\left[ \sum _{\sigma \in \Sigma _{a,a}} \Pi _\sigma \right] . \end{aligned}$$

\(\square \)

1.5 Scalar product of vectors in \(\mathcal {V}^\lambda \)

Lemma A.12

Let \(\lambda _1, \lambda _2\) be two irreducible representations of \(\mathfrak {S}_{N - 1}\), let \(\textbf{v}_1, \textbf{v}_2\) be two normalized vectors in the irreducible subspace \(\lambda _1\) and \(\lambda _2\) of \(\mathfrak {S}_{N - 1}\), and let \(1 \le k,l \le N\). Then,

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}_1|\Omega _{(0,l)} \otimes \textbf{v}_2 \rangle = {\left\{ \begin{array}{ll} d \cdot \langle \textbf{v}_1|\textbf{v}_2\rangle &{}\textrm{if } k = l \\ \left\langle \textbf{v}_1\right| \Pi _{((l - 1) \, : \, (k - 1))} \left| \textbf{v}_2\right\rangle &{}\textrm{if } k \ne l. \end{array}\right. } \end{aligned}$$

As a consequence,

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}_1|\Omega _{(0,k)} \otimes \textbf{v}_1\rangle = d \end{aligned}$$

and if \(\lambda _1\) and \(\lambda _2\) are distinct,

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}_1|\Omega _{(0,l)} \otimes \textbf{v}_2\rangle = 0. \end{aligned}$$

In particular, the vector spaces \({{\,\textrm{Span}\,}}\mathcal {V}^\lambda \), defined in Eq. (3.3), are in orthogonal direct sum.

Proof

Since both \(\textbf{v}_1\) and \(\textbf{v}_2\) are normalized, we have \(\langle \textbf{v}_i|\textbf{v}_i\rangle = 1\) for all \(i \in \{1, 2\}\). Then, from \(\langle \Omega |\Omega \rangle = d\) we have

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}_1|\Omega _{(0,k)} \otimes \textbf{v}_1\rangle&= \langle \Omega |\Omega \rangle \otimes \langle \textbf{v}_1|\textbf{v}_1\rangle \\&= d. \end{aligned}$$

When \(k \ne l\), from the equation \(\left\langle \Omega _{(0, k)} \otimes \phi \right| \left| \Omega _{(0, l)} \otimes \psi \right\rangle = \left\langle \phi \right| \Pi _{((l - 1) \,: \, (k - 1))} \left| \psi \right\rangle \) for any \(\phi , \psi \in \mathcal {H}^{\otimes (N - 1)}\), we have

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}_1|\Omega _{(0,l)} \otimes \textbf{v}_1\rangle = \left\langle \textbf{v}_1\right| \Pi _{((l - 1) \, : \, (k - 1))} \left| \textbf{v}_1\right\rangle . \end{aligned}$$

When we look at \(\lambda _1, \lambda _2\), two distinct irreducible representations of \(\mathfrak {S}_{N - 1}\), the scalar product becomes

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}_1|\Omega _{(0,l)} \otimes \textbf{v}_2 \rangle = {\left\{ \begin{array}{ll} \langle \Omega |\Omega \rangle \otimes \langle \textbf{v}_1|\textbf{v}_2\rangle &{}\text {if } k = l \\ \left\langle \textbf{v}_1\right| \Pi _{((l - 1) \, : \, (k - 1))} \left| \textbf{v}_2\right\rangle &{}\text {if } k \ne l. \end{array}\right. } \end{aligned}$$

Note that \(\Pi _{((l - 1) \,: \, (k - 1))} \left| \textbf{v}_2\right\rangle \) lives in the irreducible subspace of \(\mathfrak {S}_{N - 1}\) associated with \(\lambda _2\), since the irreducible representations are stable by any representation of permutations. Then, both scalar products are null because the irreducible spaces are orthogonal each other.\(\square \)

Remark A.13

In the particular case of the trivial representation, we have for all \(1 \le k,l \le N\) and all normalized \(\textbf{v}\) in the irreducible subspace \(\vee ^{(N - 1)}(\mathcal {H})\),

$$\begin{aligned} \langle \Omega _{(0,k)} \otimes \textbf{v}|\Omega _{(0,k)} \otimes \textbf{v}\rangle&= d \\ \langle \Omega _{(0,k)} \otimes \textbf{v}|\Omega _{(0,l)} \otimes \textbf{v}\rangle&= 1. \end{aligned}$$

From Lemma A.8, the action of the permutation operators \(\Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}}\) on the vectors \(\left| \Omega \right\rangle _{(0,k)} \otimes \left| \textbf{v}\right\rangle \), for all \(1 \le k \le N\), is

$$\begin{aligned} \Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}} \big ( \left| \Omega \right\rangle _{(0,k)} \otimes \left| \textbf{v}\right\rangle \big ) = {\left\{ \begin{array}{ll} d \cdot \left| \Omega \right\rangle _{(1,a)} \otimes \Pi ^\lambda _{\hat{\sigma },a,b,b} \left| \textbf{v}\right\rangle &{}\text { if } b = k\\ \left| \Omega \right\rangle _{(1,a)} \otimes \Pi ^\lambda _{\hat{\sigma },a,b,k} \left| \textbf{v}\right\rangle &{}\text { if } b \ne k, \end{array}\right. } \end{aligned}$$
(31)

with \(\Pi ^\lambda _\sigma = P^\lambda \, \Pi _\sigma \, P^\lambda \), and \(P^\lambda \) the projector onto the irreducible subspace \(\lambda \).

For any permutation operator \(\Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}}\), we know from Lemma A.10 that the \(V^{\lambda }\)’s form a generators of the orthogonal complement of the kernel of \(\Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}}\). However, in general they do not define a basis. We can extract a subset of linearly independent vectors to form a basis. In this basis, \(\Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}}\) can be block diagonalized due to Eq. (31), since for all irreducible representation \(\lambda \), and any vector v in \(\mathcal {V}^{\lambda }\), we have

$$\begin{aligned} \Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}} (v) \in {{\,\textrm{Span}\,}}\mathcal {V}^{\lambda }. \end{aligned}$$

Example A.14

(\(N = 3\)) Let \(x_1, x_2, x_3 \in \mathbb {R}_+\), and define

$$\begin{aligned} S = x_1 \cdot \Pi ^{\mathsf {\Gamma }}_{(0 \, 1)} + x_2 \cdot \Pi ^{\mathsf {\Gamma }}_{(0 \, 2)} + x_3 \cdot \Pi ^{\mathsf {\Gamma }}_{(0 \, 3)}, \end{aligned}$$

that is

Then, an eigenvector for the largest eigenvalue of R is

$$\begin{aligned} \chi = \beta _1 \cdot \left| \Omega \right\rangle _{(0,1)} \otimes \left| \textbf{v}\right\rangle + \beta _2 \cdot \left| \Omega \right\rangle _{(0,2)} \otimes \left| \textbf{v}\right\rangle + \beta _3 \cdot \left| \Omega \right\rangle _{(0,3)} \otimes \left| \textbf{v}\right\rangle , \end{aligned}$$

that is,

for some coefficients \(\beta _i\), and \(\textbf{v} = \displaystyle \sum _{1 \le i \le j < d} \left| ij\right\rangle \).

Appendix B: Some results in linear algebra

In the proofs, we use simple linear algebra results that are not mainstream. We gather them in this appendix for the convenience of the reader.

The first one is a consequence of [9], for which we provide a simple proof.

Lemma B.1

For all \(1 \le i, j \le n\), let \(M_{ij}\) be some matrices in \(\mathcal {M}_m\), and define the block matrix

and the matrix . We then have

$$\begin{aligned} \Vert M\Vert \le \Vert M'\Vert . \end{aligned}$$

Proof

Consider v and w two unit vectors such that \(\Vert M\Vert = \left\langle v\right| M \left| w\right\rangle \), we can decompose \(v=\displaystyle \sum \nolimits _{i=1}^n \vert v_i\rangle \otimes \vert i\rangle \) with \(v_i\in \mathbb C^m\) and we define \(v'=\displaystyle \sum \nolimits ^n_{i=1} \Vert v_i\Vert \cdot \vert i\rangle \). Since we consider the Euclidian norm, we have obviously \(\Vert v'\Vert =\Vert v\Vert =1\). We define \(w'\) similarly and then have

$$\begin{aligned} \Big | \! \left\langle v\right| M \left| w\right\rangle \! \Big |&= \bigg | \sum _{ij} \left\langle v_i\right| M_{ij} \left| w_j\right\rangle \bigg | \\&\le \sum _{ij} \Vert v_i\Vert \, \Vert M_{ij}\Vert \, \Vert w_j\Vert \quad (\text {Cauchy-Schwarz}) \\&= \sum _{ij} \Vert v_i\Vert \, \Vert w_j\Vert \, \big \langle i \Big | M^{'} \Big | j \big \rangle \\&=\big \langle v' \Big | M^{'} \Big | w' \big \rangle \\&\le \Vert v'\Vert \, \Vert w'\Vert \, || M^{'} ||\\&\le || M^{'} ||. \end{aligned}$$

which ends the proof.\(\square \)

The following lemma is useful in the paper when we relate the spectrum of \(S_x\) with the one of \(\tilde{S}\)

Lemma B.2

Let \(A\in \mathcal M\) be an operator, and let \(n\ge m\) let \(\{ f_1, \ldots , f_n \}\) be a set of vectors such that \(\mathbb C^m = {{\,\textrm{Span}\,}}\{ f_1, \ldots , f_n \}\). We suppose that for all \(i=1,\ldots ,n\) there exists \(a_{ij}\), \(j=1,\ldots ,n\) such that

$$\begin{aligned} Af_i=\sum _{i,j=1}^n a_{ij}f_j \end{aligned}$$

Denote \(\tilde{A}\) the \(n\times n\) matrix with coefficients \((a_{ij})\). Then, 

$$\begin{aligned} {{\,\textrm{Spec}\,}}{A} \subseteq {{\,\textrm{Spec}\,}}{\tilde{A}}. \end{aligned}$$

Proof

Let F be the matrix composed of the vectors \(f_i\), \(i=1,\ldots ,n\) written in column. Then, F is a \(m\times n\) matrix of rank m since \(\mathbb C^m=Vect\{ f_1, \ldots , f_n \}\). By definition of \(\tilde{A}\), we have

$$AF=F\tilde{A}$$

Furthermore, for all scalar \(\alpha \)

$$\begin{aligned} (A - \alpha \, I_m) F&= A F - \alpha \, I_m \, F \\&= F \tilde{A} - \alpha \, F\\&=F(\tilde{A}-\alpha I_n) \end{aligned}$$

Now using the fact that \(rank(AB)\le min(rank(A),rank(B))\) for all matrices A and B, if \(\alpha \in Spec A\) since \(\hbox {rank}(F)=m\)

$$\begin{aligned} \hbox {rank}(F(\tilde{A}-\alpha I_n))=\hbox {rank}((A - \alpha \, I_m) F)<m. \end{aligned}$$

Now, since \(n\ge m\), this necessarily implies that \(\hbox {rank}(\tilde{A}-\alpha I_n)<n\) which says that \(\alpha \in Spec \tilde{A}\) \(\square \)

Remark B.3

Note that the matrix \(\tilde{A}\) is not uniquely defined and depends on a choice of the way of writing \(Af_i=\sum _{i,j=1}^n a_{ij}f_j\), but the results concerning the inclusion are true for any form of the matrix \(\tilde{A}\). Of course the other inclusion is not true in general unless \(n=m\) and \(\{ f_1, \ldots , f_n \}\) is linearly independent (but this is trivial in this case).

Remark B.4

In the lemma, the condition \(Af_i=\sum _{i,j=1}^n a_{ij}f_j\) means that A leaves \(V= {{\,\textrm{Span}\,}}\{ f_1, \ldots , f_n \}\) invariant and we can rephrase this lemma by considering the restriction to A on V and then saying that \({{\,\textrm{Spec}\,}}{A_{\vert V}} \subseteq {{\,\textrm{Spec}\,}}{\tilde{A}}.\) which is what we use in the body of the paper.

Appendix C: The \(\mathcal Q\)-norm

Lemma C.1

For all \(x\in \mathbb R^N\), we have

$$\begin{aligned} \frac{1}{d}\Vert x\Vert _1\le \lambda _{\max }(S_x)\le d\Vert x\Vert _1. \end{aligned}$$

In particular, the \(\Vert \cdot \Vert _{\mathcal Q}\) quantity is nonnegative. The first inequality is saturated if and only if the matrix \(S_x\) is a (nonnegative) multiple of the identity. The second inequality is saturated if and only if x has at most one nonzero entry.

Proof

For the first inequality, note that

$$\begin{aligned} \lambda _{\max }(S_x) \ge \frac{\hbox {Tr}S_x}{d^{N+1}} = \frac{ \sum ^N_{i=1} d |x_i| \cdot d^{N-1}}{d^{N+1}} = \frac{\Vert x\Vert _1}{d}. \end{aligned}$$

The inequality above is saturated if and only if the eigenvalues of \(S_x\) are identical.

For the second inequality, we use the subadditivity of the \(\lambda _{\max }\) functional:

$$\begin{aligned} \lambda _{\max }(S_x) \le \sum ^N_{i=1} \lambda _{\max }\left( |x_i| \cdot \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \right) = \sum ^N_{i=1} d |x_i| = d\Vert x\Vert _1. \end{aligned}$$

The inequality above is saturated if and only if the matrices \(|x_i| \cdot \omega _{(0,i)} \otimes I\) have a common largest eigenvector, which can happen only if the support of x has size 0 or 1.\(\square \)

Lemma C.2

For all \(x, y \in \mathbb {R}^N_+\)\(\Vert x + y\Vert _{\mathcal {Q}} \le \Vert x\Vert _{\mathcal {Q}} + \Vert y\Vert _{\mathcal {Q}}\).

Proof

This follows from the subadditivity of the \(\lambda _{\max }\) functional: For \(x, y \in \mathbb {R}^N_+\), we have

$$\begin{aligned} \lambda _{\max } \bigg [ \sum ^{N}_{i = 1}&(x_i + y_i) \cdot \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \bigg ] \\&\le \lambda _{\max } \bigg [ \sum ^{N}_{i = 1} x_i \cdot \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \bigg ] + \lambda _{\max } \bigg [ \sum ^{N}_{i = 1} y_i \cdot \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \bigg ]. \end{aligned}$$

\(\square \)

Lemma C.3

For all \(t \in [0,1]^N\) and \(x \in \mathbb {R}^N_+\)\(\Vert t \cdot x\Vert _{\mathcal {Q}} \le \Vert x\Vert _{\mathcal {Q}}\).

Proof

Let us show that for \(0 \le x \le y\) (meaning that \(x_i\le y_i\), \(i=1,\ldots ,N\)), we have \(\Vert x\Vert _{\mathcal {Q}} \le \Vert y\Vert _{\mathcal {Q}}\). Let \(\chi = \sum _k \beta _k \left| \Omega _{(0,k)}\right\rangle \otimes \left| \textbf{v}\right\rangle \) be a normalized eigenvector corresponding to the largest eigenvalue \(\lambda _{\max }(S_x)\), from Theorem 3.13, that is

$$\begin{aligned} \lambda _{\max }(S_x) = \left\langle \chi \Big | \sum ^{N}_{i = 1} x_i \cdot \big ( \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \big ) \Big | \chi \right\rangle . \end{aligned}$$

Note that due to the Perron–Frobenius theorem, \(\beta _i > 0\) for all \(i=1,\ldots ,N\). Let us point out two facts. First using Lemma 3.6

$$\begin{aligned} \left\langle \chi \right| \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \left| \chi \right\rangle&= \sum ^N_{k = 1} \sum ^N_{l = 1} \beta _k \beta _l \cdot \left\langle \Omega _{(0,k)} \otimes \textbf{v}\right| \big ( \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \big ) \left| \Omega _{(0,l)} \otimes \textbf{v}\right\rangle \\&= \sum ^N_{k = 1} \beta _k \cdot \big \langle \Omega _{(0,k)} \otimes \textbf{v} \Big | d \beta _i \cdot \Omega _{(0,i)} \otimes \textbf{v} + \sum ^N_{\begin{array}{c} l = 1 \\ l \ne i \end{array}} \beta _l \cdot \Omega _{(0,i)} \otimes \textbf{v} \big \rangle \\&= \sum ^N_{k = 1} \beta _k \cdot \big \langle \Omega _{(0,k)} \otimes \textbf{v} \Big | (d - 1) \beta _i \cdot \Omega _{(0,i)} \otimes \textbf{v} + \sum ^N_{l = 1} \beta _l \cdot \Omega _{(0,i)} \otimes \textbf{v} \big \rangle \\&= {\bigg ( (d - 1) \beta _i + \sum _k \beta _k \bigg )}^2 \end{aligned}$$

Second

$$\begin{aligned} 1&=\langle \chi |)\rangle \\&= \sum _{i\ne j}\beta _i\beta _j+d\sum _i\beta _i^2\\&= \left( \sum _{i=1}^N\beta _i\right) ^2+(d-1)\sum _{i=1}^N\beta _i^2 \end{aligned}$$

Now using that \(\left( \sum _{i=1}^N\beta _i\right) ^2\ge \sum _{i=1}^N\beta _i^2\) we get that \(1\le d\left( \sum _{i=1}^N\beta _i\right) ^2\) then since \(\beta _i\ge 0\)

$$\begin{aligned} \left\langle \chi \right| \omega _{(0,i)} \otimes I^{\otimes (N - 1)} \left| \chi \right\rangle \ge \left( \sum _{i=1}^N\beta _i\right) ^2\ge \frac{1}{d} \end{aligned}$$

This way since \(y\ge x\), we have that \(\lambda _{max}(S_y)\ge \langle \chi ,S_y\chi \rangle \) then

$$\begin{aligned} \lambda _{max}(S_y)-\lambda _{max}(S_x)&\ge \langle \chi \vert S_y-S_x\vert \chi \rangle \\&=\sum _{i=1}^N(y_i-x_i)\langle \chi \vert \omega _{(0,i)} \otimes I^{\otimes (N - 1)}\vert \chi \rangle \\&=\sum _{i=1}^N(y_i-x_i){\bigg ( (d - 1) \beta _i + \sum _k \beta _k \bigg )}^2\\&\ge \sum _{i=1}^N(y_i-x_i)\frac{1}{d} \end{aligned}$$

Then,

$$\begin{aligned} \Vert y\Vert _{\mathcal {Q}}=\lambda _{max}(S_y)-\frac{1}{d} \Vert y\Vert \ge \lambda _{max}(S_x)-\frac{1}{d} \Vert x\Vert =\Vert x\Vert _{\mathcal {Q}} \end{aligned}$$

which was the desired result.\(\square \)

Appendix D: Optimal cloning

Lemma D.1

Let some positive real numbers \({(\beta _i)}_{1 \le i \le N}\). The operator

$$\begin{aligned} \widetilde{C}_{T_\beta }=\sum _{\begin{array}{c} 1 \le a,b \le N \\ \sigma \in \Sigma _{a,b} \end{array}} \frac{\beta _a \beta _b}{(N - 1)!} \Pi _\sigma ^{\mathsf {\Gamma }} \end{aligned}$$

is an orthogonal projection if and only if the condition

$$\begin{aligned} (d - 1) \sum ^N_{i = 1} \beta ^2_i + {\bigg ( \sum ^N_{i = 1} \beta _i \bigg )}^2 = 1 \end{aligned}$$
(32)

is satisfied. In this case, the operator

$$\begin{aligned} C_{T_\beta } = \frac{d}{\hbox {Tr}P^+_{\mathfrak {S}_N}} \frac{N + d - 1}{N} \cdot \widetilde{C}_{T_\beta }, \end{aligned}$$

is a positive operator, such that \(\hbox {Tr}_{\scriptscriptstyle [\![ 1,N]\!]} C_{T_\beta } = I\)

Proof

Let some positive real numbers \({(\beta _i)}_{1 \le i \le N}\), then

$$\begin{aligned} {\big ( \widetilde{C}_{T_\beta } \big )}^2&= \sum _{\begin{array}{c} 1 \le a,b \le N \\ 1 \le c,d \le N \end{array}} \, \sum _{\begin{array}{c} \sigma \in \Sigma _{a,b} \\ \tau \in \Sigma _{c,d} \end{array}} \frac{\beta _a \beta _b \beta _c \beta _d}{{(N-1)!}^2} \, \Pi _\sigma ^{\mathsf {\Gamma }} \Pi _\tau ^{\mathsf {\Gamma }} \\&= \sum _{1 \le a,d \le N} \left[ d \sum _{\begin{array}{c} 1 \le b,c \le N \\ b = c \end{array}} \sum _{\sigma \in \Sigma _{a,d}} \frac{\beta _a \beta _b \beta _c \beta _d}{{(N-1)!}^2} \, \Pi _\sigma ^{\mathsf {\Gamma }} + \sum _{\begin{array}{c} 1 \le b,c \le N \\ b \ne c \end{array}} \sum _{\sigma \in \Sigma _{a,d}} \frac{\beta _a \beta _b \beta _c \beta _d}{{(N-1)!}^2} \, \Pi _\sigma ^{\mathsf {\Gamma }} \right] . \end{aligned}$$

The projection condition \({\big ( \widetilde{C}_{T_\beta } \big )}^2 = \widetilde{C}_{T_\beta }\) is equivalent to

$$\begin{aligned} \beta _a \beta _b&= d \sum ^N_{i = 1} \beta _a \beta _i \beta _i \beta _b + \sum _{\begin{array}{c} 1 \le i,j \le N \\ i \ne j \end{array}} \beta _a \beta _i \beta _j \beta _b \\&= (d - 1) \sum ^N_{i = 1} \beta _a \beta _i \beta _i \beta _b + \sum _{1 \le i,j \le N} \beta _a \beta _i \beta _j \beta _b \\&= \beta _a \bigg ( (d - 1) \sum ^N_{i = 1} \beta _i \beta _i + \sum _{1 \le i,j \le N} \beta _i \beta _j \bigg ) \beta _b \\&= \beta _a \bigg ( (d - 1) \sum _{i=1}^N\ \beta ^2_i + {\bigg ( \sum _i \beta _i \bigg )}^2 \bigg ) \beta _b. \end{aligned}$$

Since (17) is assumed to be satisfied, then \(\widetilde{C}_{T_\beta }\) is an orthogonal projection and hence a positive operator. This way \(C_{T_\beta }\) is positive. We have now from Lemma A.11

$$\begin{aligned} \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \big ( \widetilde{C}_{T_\beta } \big )&= \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \bigg [ \sum _{\begin{array}{c} 1 \le a,b \le N \\ \sigma \in \Sigma _{a,b} \end{array}} \frac{\beta _a \beta _b}{(N - 1)!} \sigma ^{\mathsf {\Gamma }} \bigg ] \\&= \hbox {Tr}\big [ P^+_{\mathfrak {S}_{N - 1}} \Big ] \bigg ( \sum _i \beta _i \beta _i + \frac{1}{d} \sum _{i \ne j} \beta _i \beta _j \bigg ) \cdot I \\&= \frac{\hbox {Tr}P^+_{\mathfrak {S}_{N - 1}}}{d} \bigg [ (d - 1) \sum _i \beta ^2_i + {\bigg ( \sum _i \beta _i \bigg )}^2 \bigg ] \cdot I \\&= \frac{\hbox {Tr}P^+_{\mathfrak {S}_{N - 1}}}{d} \cdot I \end{aligned}$$

Then,

$$\begin{aligned} \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} (C_{T_\beta })&= \hbox {Tr}_{\scriptscriptstyle [\![ 1,N ]\!]} \big [ \frac{d}{\hbox {Tr}P^+_{\mathfrak {S}_N}} \frac{N + d - 1}{N} \cdot \widetilde{C}_{T_\beta } \Big ] \\&= \frac{\hbox {Tr}P^+_{\mathfrak {S}_{N - 1}}}{\hbox {Tr}P^+_{\mathfrak {S}_N}} \frac{N + d - 1}{N} \cdot I \\&= I, \end{aligned}$$

where the last equation comes from \(\hbox {Tr}\big [ P^+_{\mathfrak {S}_{N - 1}} \Big ] = \frac{N}{N + d - 1} \hbox {Tr}\big [ P^+_{\mathfrak {S}_{N}} \Big ]\).\(\square \)

Lemma D.2

For any \(1 \le a,b \le N\), and any \(\sigma \in \Sigma _{a,b}\), the permutation operator \(\sigma ^{\mathsf {\Gamma }}\) is the Choi matrix a linear map \(T_{\mu ,\nu }: \mathcal {M}_d \rightarrow {(\mathcal {M}_d)}^{\otimes N}\) defined by

$$\begin{aligned} T_{\mu ,\nu }(X) = \Pi _\mu \big ( X \otimes I^{\otimes (N - 1)} \big ) \Pi _\nu \end{aligned}$$

for some permutations \(\mu \) and \(\nu \) in \(\mathfrak {S}_N\) such that \(\mu (0) = a - 1\) and \(\nu (b - 1) = 0\).

As a consequence for some fixed \(1 \le a,b \le N\),

$$\begin{aligned} \sum _{\sigma \in \Sigma _{a,b}} \Pi _\sigma ^{\mathsf {\Gamma }} = \frac{1}{(N - 1)!} C_{T_{a,b}} \end{aligned}$$

where

$$\begin{aligned} T_{a,b}(X) = \sum _{\begin{array}{c} \scriptstyle \mu ,\nu \scriptstyle \in \mathfrak {S}_N \\ \scriptstyle \mu (0) \scriptstyle = a - 1 \\ \scriptstyle \nu (b - 1) \scriptstyle = 0 \end{array}} \Pi _\mu \big ( X \otimes I^{\otimes (N - 1)} \big ) \Pi _\nu . \end{aligned}$$

Proof

Let \(1 \le a,b \le N\), and \(\sigma \in \Sigma _{a,b}\). Then from Lemma A.5, we have

$$\begin{aligned} \Pi _\sigma ^{\mathsf {\Gamma }} = \Pi _{(1 \, a)} \, (\omega \otimes \Pi _{\hat{\sigma }}) \; \Pi _{(1 \, b)}, \end{aligned}$$

such that on \(X \in \mathcal {M}_d\), the partial trace yields

And the result holds for \(\Pi _\mu = \Pi _{(0 \, (a - 1))}\) and \(\Pi _\nu = (I \otimes \Pi _{\hat{\sigma }}) \circ \Pi _{(0 \, (b - 1))}\). \(\square \)

Lemma D.3

Let an orthogonal projector \(\widetilde{C}_{T_\beta }\) for some positive real number \({(\beta _i)}_{1 \le i \le N}\) satisfying Eq. (17). Then for any \(1 \le i \le N\), we have

$$\begin{aligned} \widetilde{C}_{T_\beta } \, \omega _{(0,i)} = \frac{1}{(N - 1)!} \sum _{\begin{array}{c} 1 \le a \le N \\ \sigma \in \Sigma _{a,i} \end{array}} \beta _a \bigg ( (d - 1) \beta _i + \sum _{1 \le b \le N} \beta _b \bigg ) \Pi _\sigma ^{\mathsf {\Gamma }}. \end{aligned}$$

Let Choi matrix \(C_{T_\beta }\) for some positive real numbers \({(\beta _i)}_{1 \le i \le N}\) satisfying Eq. (17). Then for any \(1 \le i \le N\), we have

$$\begin{aligned} \hbox {Tr}\big [ C_{T_\beta } \, \omega _{(0,i)} \Big ] = d {\bigg ( (d - 1) \beta _i + \sum ^N_{j = 1} \beta _j \bigg )}^2. \end{aligned}$$

Proof

Since we are summing on all permutations \(\Pi ^{\mathsf {\Gamma }}_{\sigma _{a,b}}\), a direct calculation gives us

$$\begin{aligned} \widetilde{C}_{T_\beta } \, \omega _{(0,i)}&= \frac{1}{(N - 1)!} \sum _{\begin{array}{c} 1 \le a,b \le N \\ \sigma \in \Sigma _{a,b} \end{array}} \beta _a \beta _b \cdot \Pi _\sigma ^{\mathsf {\Gamma }} \, \omega _{(0,i)} \\&= \frac{1}{(N - 1)!} \sum _{1 \le a \le N} \beta _a \bigg ( (d - 1) \sum _{\sigma \in \Sigma _{a,i}} \beta _i \cdot \sigma ^{\mathsf {\Gamma }} + \sum _{\begin{array}{c} 1 \le b \le N \\ \sigma \in \Sigma _{a,i} \end{array}} \beta _b \cdot \Pi _\sigma ^{\mathsf {\Gamma }} \bigg ) \\&= \frac{1}{(N - 1)!} \sum _{\begin{array}{c} 1 \le a \le N \\ \sigma \in \Sigma _{a,i} \end{array}} \beta _a \bigg ( (d - 1) \beta _i + \sum _{1 \le b \le N} \beta _b \bigg ) \Pi _\sigma ^{\mathsf {\Gamma }}. \end{aligned}$$

And then we look at the action of the Choi matrix \(C_{T_\beta }\) onto some \(\omega _{(0,i)}\) and take the trace.

Lemma A.11 yields

$$\begin{aligned} \hbox {Tr}\big [ C_{T_\beta } \, \omega _{(0,i)} \Big ]&= \frac{d (N + d - 1)}{N \, \hbox {Tr}\! \big [ P^+_{\mathfrak {S}_N} \Big ]} \sum ^N_{a = 1} \frac{\beta _a}{(N - 1!)} \bigg ( (d - 1) \beta _i + \sum ^N_{b = 1} \beta _b \bigg ) \hbox {Tr}\bigg [ \sum _{\sigma \in \Sigma _{a,i}} \Pi _\sigma ^{\mathsf {\Gamma }} \bigg ] \\&= \frac{d (N + d - 1)}{N \, \hbox {Tr}\! \big [ P^+_{\mathfrak {S}_N} \Big ]} \bigg ( (d - 1) \beta _i + \sum ^N_{b = 1} \beta _b \bigg ) \bigg ( (d - 1) \beta _i + \sum ^N_{a = 1} \beta _a \bigg ) \hbox {Tr}\big [ P^+_{\mathfrak {S}_{N - 1}} \Big ] \\&= \frac{d (N + d - 1)}{N \, \hbox {Tr}\! \big [ P^+_{\mathfrak {S}_N} \Big ]} {\bigg ( (d - 1) \beta _i + \sum ^N_{j = 1} \beta _j \bigg )}^2 \hbox {Tr}\big [ P^+_{\mathfrak {S}_{N - 1}} \Big ] \\&= d {\bigg ( (d - 1) \beta _i + \sum ^N_{j = 1} \beta _j \bigg )}^2. \end{aligned}$$

\(\square \)

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Nechita, I., Pellegrini, C. & Rochette, D. The asymmetric quantum cloning region. Lett Math Phys 113, 74 (2023). https://doi.org/10.1007/s11005-023-01694-8

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s11005-023-01694-8

Mathematics Subject Classification

Navigation