This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Observation of the Knot Topology of Non-Hermitian Systems in a Single Spin

Yang Wu    Yunhan Wang    Xiangyu Ye    Wenquan Liu CAS Key Laboratory of Microscale Magnetic Resonance and School of Physical Sciences, University of Science and Technology of China, Hefei 230026, China CAS Center for Excellence in Quantum Information and Quantum Physics, University of Science and Technology of China, Hefei 230026, China    Chang-Kui Duan    Ya Wang    Xing Rong [email protected] CAS Key Laboratory of Microscale Magnetic Resonance and School of Physical Sciences, University of Science and Technology of China, Hefei 230026, China CAS Center for Excellence in Quantum Information and Quantum Physics, University of Science and Technology of China, Hefei 230026, China Hefei National Laboratory, University of Science and Technology of China, Hefei 230088, China    Jiangfeng Du [email protected] CAS Key Laboratory of Microscale Magnetic Resonance and School of Physical Sciences, University of Science and Technology of China, Hefei 230026, China CAS Center for Excellence in Quantum Information and Quantum Physics, University of Science and Technology of China, Hefei 230026, China Hefei National Laboratory, University of Science and Technology of China, Hefei 230088, China School of Physics, Zhejiang University, Hangzhou 310027, China
Abstract

The non-Hermiticity of the system gives rise to distinct knot topology that has no Hermitian counterpart. Here, we report a comprehensive study of the knot topology in gapped non-Hermitian systems based on the universal dilation method with a long coherence time nitrogen-vacancy center in a 12{}^{\text{12}}C isotope purified diamond. Both the braiding patterns of energy bands and the eigenstate topology are revealed. Furthermore, the global biorthogonal Berry phase related to the eigenstate topology has been successfully observed, which identifies the topological invariance for the non-Hermitian system. Our method paves the way for further exploration of the interplay among band braiding, eigenstate topology and symmetries in non-Hermitian quantum systems.

The non-Hermitian (NH) Hamiltonian has drawn intensive attention RMP_Bergholtz ; PRL_NHtopo1 ; PRX_NHtopo1 ; PRX_NHtopo2 ; Science_NHtopo ; PRL_NHtopo2 ; PRA_NHtopo due to its application in photonic Science_NHApli1 ; NP_NHApli ; Science_NHApli2 ; Science_NHApli3 ; Nature_PhotoNH ; Science_PhotoNH1 ; PRL_PhotoNH1 ; PRL_PhotoNH2 ; PRL_PhotoNH3 ; NaturePhotonics_PhotoNH ; NP_PhotoNH and acoustic NHAcous_Romain ; NHAcous_Tuo ; NHAcous_Zhi systems with gain and loss, open quantum systems JPA_OpenNH ; Nature_OpenNH ; PRL_OpenNHCar ; NP_OpenNHDiehl and condensed matter systems of quasi-particles with finite lifetime Arxiv_CondNH ; PRB_CondNH1 ; PRB_CondNH2 ; PRL_CondNH . In contrast to Hermitian systems, topological structures that are exclusive to NH systems give birth to many intriguing phenomena like NH bulk boundary correspondence PRL_Kai ; PRL_Dan and NH skin effects PRL_Nobuyuki . The emergence of NH nodal phases and gapped phases as well as their interplay with symmetry also adds new ingredients for NH topological band theory RMP_Bergholtz . Understanding the topology introduced by NH physics helps one to explore topologically robust quantities and non-trivial edge states. All these uncommon phenomena provide insights into NH systems.

More recently, researchers have found a framework based on homotopy theory PRB_Fan ; PRL_Haiping ; PRB_ZhiLi to classify the non-Hermitian topological phases. Compared with the K-theory PRX_NHtopo1 , this classification method does not need to assume specific symmetries and can reveal more topological invariants. To be precise, the Hamiltonian can be regarded as a mapping from the Brillouin zone to the space of energy bands and eigenstates. Thus the classification is done by finding all topologically non-equivalent mappings. Each class of mapping corresponds to a topological phase with a specific topological invariant. The non-Hermiticity of the Hamiltonian can generate complex eigenvalues, which can constitute the knot structures. The whole classification set of NH systems with knot topology is decomposed into several sectors, based on the braiding of eigenvalues, and each sector can be further classified with the eigenstate topology PRB_ZhiLi .

There are several experimental studies recently aiming at revealing the knot topology of the NH systems. For example, the braiding structure of energy bands has been observed in optical Nature_Fan , mechanical Nature_Harris and trapped ion PRL_Cao systems. A comprehensive demonstration of the knot topology, especially the essential information from the eigenstate topology requires coherent evolution under the NH Hamiltonian. It is difficult to realize NH Hamiltonians in quantum systems since closed systems are generally governed by Hermitian Hamiltonians. Apart from the difficulty in realizing a NH Hamiltonian, the inevitable decoherence of the quantum system tends to destroy the coherent evolution. In order to overcome these difficulties, we applied the universal dilation method Science_YW ; PRL_Wengang ; PRL_Uwe ; PRL_Kohei to realize the NH Hamiltonians and fabricated a 99.999% 12{}^{\text{12}}C isotope purified sample in which the decoherence had been significantly suppressed. Taking the one-dimensional (1D) NH lattice model as an example, we experimentally studied the knot topology in the NH Hamiltonian by measuring not only its energy eigenvalues, but also the corresponding eigenstates to reveal the global Berry phase. Compared with the NH topology shown in previous work PRL_Wengang on the NV center, which stems from encircling varying numbers of exceptional points, the knot topology investigated in our work arises from both the braiding of all energy bands and the eigenstate topology. Our platform is capable of dealing with other models and our results shed light on the complexity of topological features aroused by non-Hermiticity.

Here we consider a two-band NH 1D lattice model (Fig. 1(a)) whose Hamiltonian may be written as

(m)=j[Γ0+ajbj+Γ0ajbj+n=1m(Γ1n+ajbj+n+Γ1najbj+n+Γ2n+aj+nbj+Γ2naj+nbj)],\begin{split}&\mathcal{H}^{(m)}=\sum_{j}[\Gamma^{0+}a_{j}b_{j}^{\dagger}+\Gamma^{0-}a_{j}^{\dagger}b_{j}+\\ &\sum_{n=1}^{m}(\Gamma^{n+}_{1}a_{j}b_{j+n}^{\dagger}+\Gamma^{n-}_{1}a_{j}^{\dagger}b_{j+n}+\Gamma^{n+}_{2}a_{j+n}^{\dagger}b_{j}+\Gamma_{2}^{n-}a_{j+n}b_{j}^{\dagger})],\end{split} (1)

where aj(bj)a_{j}\ (b_{j}) is the annihilation operator of sublattice a(b)a\ (b) on the jjth site, and the Γ\Gamma’s are complex numbers representing the coupling strengths, i.e., Γ0±\Gamma^{0\pm} is the on-site interaction and Γ1(2)n±\Gamma_{1(2)}^{n\pm} is the hopping between sublattice a(b)a\ (b) at site jj and lattice b(a)b\ (a) at site j+nj+n. Under the periodic boundary condition, the corresponding momentum space Hamiltonian is

H(m)(k)=[0Γ0Γ0+0]+\displaystyle H^{(m)}(k)=\left[\begin{array}[]{ccc}0&\Gamma^{0-}\\ \Gamma^{0+}&0\end{array}\right]+ (2)
n=1m[0Γ1neink+Γ2n+einkΓ1n+eink+Γ2neink0].\displaystyle\sum_{n=1}^{m}\left[\begin{array}[]{ccc}0&\Gamma_{1}^{n-}e^{ink}+\Gamma_{2}^{n+}e^{-ink}\\ \Gamma_{1}^{n+}e^{-ink}+\Gamma_{2}^{n-}e^{ink}&0\end{array}\right].

This model holds plenty of topological phases under different parameter configurations that ensure bands being separable. For m=1m=1, i.e., the nearest coupling case, the system may be in the unlink phase or the unknot phase. Because k=0k=0 and 2π2\pi are equivalent due to the periodic condition, the energy bands form closed curves. When the parameters are set as Γ0=0.45,\Gamma^{0-}=-0.45, Γ0+=0.79,\Gamma^{0+}=0.79, Γ11=0.30i,\Gamma^{1-}_{1}=-0.30i, Γ21=0.08i,\Gamma^{1-}_{2}=0.08i, and Γ1,21+=0\Gamma^{1+}_{1,2}=0, the non-Hermitian system is in the unlink phase, where the two circles are unrelated, as shown by Fig. 1(b). If the parameters are changed to Γ0=0.21,\Gamma^{0-}=-0.21, and Γ0+=0.70\Gamma^{0+}=0.70, the system will be in the unknot phase. There is only one circle because the two bands exchange (Fig. 1(c)). For m=2m=2, the next nearest coupling enriches the topological phases in addition to the unlink and unknot. When the parameters are set as Γ0=0.04,\Gamma^{0-}=0.04, Γ0+=0.49,\Gamma^{0+}=0.49, Γ11=Γ21+=0.13i,\Gamma^{1-}_{1}=\Gamma^{1+}_{2}=-0.13i, Γ21=Γ11+=0.02i,\Gamma^{1-}_{2}=\Gamma^{1+}_{1}=0.02i, Γ12=0.58i,\Gamma^{2-}_{1}=-0.58i, Γ22+=0.21i,\Gamma^{2+}_{2}=-0.21i, Γ12+=0.03i,\Gamma^{2+}_{1}=0.03i, and Γ22=0.09i\Gamma^{2-}_{2}=0.09i, the two eigenvalues form a non-trivial braiding structure. The energy bands braid around each other exactly once and this gives the Hopf link phase (Fig. 1(d)). The eigenstates also exhibit behaviors similar to those of the corresponding energy bands as shown in Fig. 1(e), 1(f) and 1(g). For larger mm, the energy bands may braid more times, which leads to more phases. The classification based on homotopy theory has shown that all phases of this model are described by the braiding group B(2)B(2) PRB_Fan ; PRB_ZhiLi .

Refer to caption
Figure 1: The knot topology of eigenvalues and eigenstates in a 1D NH lattice model. (a) Part of the lattice with on-site (green), nearest (yellow) and next nearest (red) interaction. (b-d) Band structure and the corresponding knot diagrams of three phases under different parameter configurations. The projection of eigenvalues on the complex plane (dashed lines) shows different topological structures. (b) Two completely separated bands, which are similar to an insulator in the Hermitian case. (c) Two bands exchange and the projection lines merge into a whole circle. (d) Two bands encircle each other and become the Hopf link. (e-g) The eigenstates show the same pattern as the corresponding energy bands.

We now study the topological phases of our model and the corresponding topological invariants in a quantum system based on the eigenvalues and eigenstates. Both the eigenvalues and the eigenstates can be obtained from the evolution under the NH Hamiltonian Hs(k)=H(m)(k)H_{s}(k)=H^{(m)}(k). Let |ψ1,2(k)\ket{\psi_{1,2}(k)} (the kk dependence is omitted for simplicity in the following) be the eigenstates with complex eigenvalues E1,2=E1,2r+iE1,2iE_{1,2}=E^{r}_{1,2}+iE^{i}_{1,2}, where EnrE^{r}_{n} and EniE^{i}_{n} are real and imaginary parts of EnE_{n}, respectively. For any initial state written as |ψ(0)=c1|ψ1+c2|ψ2\ket{\psi(0)}=c_{1}\ket{\psi_{1}}+c_{2}\ket{\psi_{2}}, the evolution governed by Hs(k)H_{s}(k) gives |ψ(t)c1eiE1rt+E1it|ψ1+c2eiE2rt+E2it|ψ2\ket{\psi(t)}\propto c_{1}e^{-iE^{r}_{1}t+E^{i}_{1}t}\ket{\psi_{1}}+c_{2}e^{-iE^{r}_{2}t+E^{i}_{2}t}\ket{\psi_{2}}. Without loss of generality, we assume E1i>E2iE^{i}_{1}>E^{i}_{2}. The eigenvalues can be extracted from the time evolution of the populations of |ψ1,2\ket{\psi_{1,2}}, and the steady state of NH evolution will be the eigenstate |ψ1\ket{\psi_{1}} since E1iE^{i}_{1} is larger than E2iE^{i}_{2}. By implementing evolution governed by Hs-H_{s}, the eigenstate |ψ2\ket{\psi_{2}} can be obtained, which is due to E2i>E1i-E^{i}_{2}>-E^{i}_{1} for Hs-H_{s}.

The evolution under the NH Hamiltonian Hs(k)H_{s}(k) can be realized based on the universal dilation method. The state of the system |ψ(t)\ket{\psi(t)} needs to evolve as it|ψ(t)=Hs(k)|ψ(t)i\partial_{t}\ket{\psi(t)}=H_{s}(k)\ket{\psi(t)}. By introducing an ancilla, the NH evolution can be realized in a subspace while the total Hamiltonian HtotH_{\rm tot} is Hermitian. The initial state that reads |0s|a+η(0)|0s|+a\ket{0}_{s}\ket{-}_{a}+\eta(0)\ket{0}_{s}\ket{+}_{a} evolves to |ψ(t)s|a+η(t)|ψ(t)s|+a\ket{\psi(t)}_{s}\ket{-}_{a}+\eta(t)\ket{\psi(t)}_{s}\ket{+}_{a} under the Hamiltonian HtotH_{\rm tot}, where |±\ket{\pm} are eigenstates of σy\sigma_{y} and η(t)\eta(t) is a properly chosen time-dependent operator. Thus in the |a\ket{-}_{a} subspace, apart from a normalization constant, the evolution is strictly governed by Hs(k)H_{s}(k).

We use a single nitrogen-vacancy (NV) center in diamond to experimentally realize the momentum space NH Hamiltonian for different phases with k[0,2π]k\in[0,2\pi] (see Appendix A for details of the experimental setup). The NV center is a type of point defect in diamond that is composed of a nitrogen atom and a neighbor vacancy as shown in Fig. 2(a). The ground-state of the NV center is a triplet state with an electronic spin S=1S=1 that interacts with the nuclear spin I=1I=1 of the substitutional 14{}^{\text{14}}N. We construct the dilated Hamiltonian HtotH_{\rm tot} in the subspace spanned by the states |mS,mI=|0,1,|0,0,|1,1\ket{m_{S},m_{I}}=\ket{0,1},\ket{0,0},\ket{-1,1}, and |1,0\ket{-1,0} as shown in Fig. 2(b). The energy levels are relabeled as |1e|1n,|1e|0n,|0e|1n\ket{1}_{e}\ket{1}_{n},\ket{1}_{e}\ket{0}_{n},\ket{0}_{e}\ket{1}_{n}, and |0e|0n\ket{0}_{e}\ket{0}_{n}, respectively. The electron spin is chosen to be the system and nuclear spin serves as the ancilla.

Refer to caption
Figure 2: Realization of the NH Hamiltonian in the NV center. (a) Simplified atomic structure of an NV center. Light cyan balls are carbon atoms, the yellow ball is a nitrogen atom and the purple one is a vacancy. (b) Four used energy levels of the NV ground state with hyperfine and nuclear quadrupolar interaction. Microwave pulses with different frequencies can be applied to selectively drive the electron spin in the |mI=1,0\ket{m_{I}=1,0} subspace. Similarly, the radio-frequency (RF) field is used to selectively control the nuclear spin. (c) Dephasing time of the NV center in our experiment. Fitting of luminescence in the Ramsey experiment shows that T2=78(7)μT_{2}^{\star}=78(7)\ \upmus, which is long enough to maintain coherence in our experiment. (d) The pulse sequence of our experiment. State preparation is realized by laser pumping and an RF π/2\pi/2 pulse Rϕ(π/2)R_{\phi}(\pi/2) with the phase ϕ\phi depending on η0\eta_{0}. The evolution under HtotH_{tot} is realized by two selective MW pulses with time-dependent amplitude, frequency and phase. Experimental results for the Hopf link phase; k=0.6πk=0.6\pi are shown in panels (e) and (f). (e) Dynamics of the population. (f) Real and imaginary parts of the MLE results for the final state of evolution. Orange bars are theoretical predictions and blue bars with error bars are experimental results.

The pulse sequence used to realize the evolution under the NH Hamiltonian is shown in Fig. 2(d). The external static magnetic field was set to be 506 G. The state of the NV center was polarized to |0e|1n\ket{0}_{e}\ket{1}_{n} by 532-nm laser pulses Polarization . After polarization, the initial state is prepared by the RF π/2\pi/2 pulse Rϕ(π/2)R_{\phi}(\pi/2), where the rotation axis lies in the XY plane and ϕ=\phi=atan[(η021)/2η0]+π/2[(\eta_{0}^{2}-1)/2\eta_{0}]+\pi/2 is the angle between the rotation axis and the xx axis. Here we have chosen η(0)=η0I\eta(0)=\eta_{0}I. The evolution under HtotH_{\rm tot} is realized by applying two selective microwave (MW) pulses, and the coherent evolution should be long enough to drive the system to a steady state. To this end, HsH_{s} is multiplied by an overall coefficient λ\lambda to speed up the evolution and preserve the coherence. The strength of the MW pulses is proportional to the value of λ\lambda. However, too strong MW pulses may cause strong crosstalk between different subspaces. One possible solution is to use a strongly coupled C13{}^{13}\text{C} nuclear spin. For the NV centers in diamond with 12C natural abundance, if one chooses λ\lambda to maintain coherence, the system cannot reach the steady state because of crosstalk. Or if one overcomes the crosstalk by using weak MW pulses, the required evolution time is so long that the coherence will greatly decrease during evolution (see Appendix C). Therefore, it is challenging to study the knot topological features of NH systems with a 12C natural abundance diamond. To address this issue, we synthesized a diamond with 99.999% 12{}^{\text{12}}C isotope abundance by the chemical vapor deposition method. With this sample, the dephasing time of the NV center utilized in our experiment is T2=78(7)T_{2}^{\star}=78(7) μ\upmus, as shown in Fig. 2(c). Such a long coherence time enables us to realize evolution under the NH Hamiltonian during which the coherence is preserved and the crosstalk is suppressed. To realize the evolution under HtotH_{\rm tot}, we choose a proper rotating frame and use two selective microwave sequences with time-dependent amplitude and phase (see Appendix B). The eigenvalues are extracted from the population information by setting Ri=IR_{i}=I in the measurement sequence. The eigenstates of the NH Hamiltonian are reconstructed by applying Ri=I,Ry(π/2)R_{i}=I,R_{-y}(\pi/2) and Rx(π/2)R_{-x}(\pi/2) to the final state of evolution. Before readout of the photoluminescence ratio, we apply an RF π/2\pi/2 pulse, which changes the basis of the ancilla from |±n\ket{\pm}_{n} to |0,1n\ket{0,1}_{n}. Finally, selective π\pi pulses are applied to reverse the population of electrons or nuclear spin to get a set of equations that relate photoluminescence intensities to populations of each level. Solving this equation, the populations are obtained.

Both eigenvalues and eigenstates can be obtained from the measurement sequences. When Ri=IR_{i}=I, we obtain a set of populations at different times by varying the evolution time under HtotH_{\rm tot}. Renormalizing the population as P1=P|1e|1n/(P|1e|1n+P|0e|1n)P_{1}=P_{\ket{1}_{e}\ket{1}_{n}}/(P_{\ket{1}_{e}\ket{1}_{n}}+P_{\ket{0}_{e}\ket{1}_{n}}) gives the evolution of P1P_{1} under the NH Hamiltonian. The corresponding quasi-momentum kk can be extracted by fitting P1P_{1} under fixed model parameters. Figure 2(e) shows the result for the Hopf link phase with k=0.6πk=0.6\pi. The population evolution agrees with the theoretical prediction and gives kfit=0.59(7)πk_{\rm fit}=0.59(7)\pi. The eigenvalues can thus be computed from kfitk_{\rm fit} and model parameters. By measuring the populations under different bases, we obtain the expectation σx,y,z\braket{\sigma_{x,y,z}}. Then we can reconstruct the steady state by ρ=(I+σσ)/2\rho=(I+\braket{\vec{\sigma}}\vec{\sigma})/2, but the direct result may give a mixed state or even an unphysical state. Thus, we use the maximum likelihood estimation (MLE) to obtain a pure state close to the eigenstate (see Appendix D). Figure 2(f) shows the real and imaginary parts of the measured state and theoretical results for k=0.6πk=0.6\pi in the Hopf link phase. Here ρ1=|ψ1ψ1|\rho_{1}=\ket{\psi_{1}}\bra{\psi_{1}} is the density matrix corresponding to |ψ1\ket{\psi_{1}}. The fidelity is 99% for this state. All other states are obtained by following the same procedure and the fidelities are all higher than 97%.

The experimentally measured eigenvalues of different phases for various kk are plotted in Fig. 3. The B(2)B(2) braiding behavior is characterized by the following definition of the winding number Nature_Fan :

ν=02πdk2πiddklnDet{H(m)(k)12Tr[H(m)(k)]}.\nu=\int_{0}^{2\pi}\frac{dk}{2\pi i}\frac{d}{dk}\text{lnDet}\left\{H^{(m)}(k)-\frac{1}{2}\text{Tr}[H^{(m)}(k)]\right\}. (3)

This number ν\nu reflects how many times the energy bands are braided. The term Tr[H(m)(k)]/2\text{Tr}[H^{(m)}(k)]/2 here can eliminate the dependence on the choice of reference energy Nature_Fan . So, the winding number defined in Eq. (3) only captures the mutual braiding of the eigenvalues under investigation. Every element of B(2)B(2) can be expressed as τn\tau^{n}, where nn is an integer and τ\tau is the generator. The one-to-one correspondence of ν=n\nu=n exactly describes the B(2)B(2) behavior. When m=1,Γ0=0.45,Γ0+=0.79,Γ11=0.30i,Γ21=0.08i,m=1,\Gamma^{0-}=-0.45,\Gamma^{0+}=0.79,\Gamma^{1-}_{1}=-0.30i,\Gamma^{1-}_{2}=0.08i, and Γ1,21+=0\Gamma^{1+}_{1,2}=0, the system is in the unlink phase (Fig. 3(a)). The eigenvalues form two separated curves and ν=0\nu=0 in this phase. The two circles can be separated by a line, Im(E)=0\text{Im}(E)=0 for example. This is just like a normal insulator in the Hermitian case. If we set the parameters as Γ0=0.21\Gamma^{0-}=-0.21 and Γ0+=0.70\Gamma^{0+}=0.70, the system will be in the unknot phase, as shown by Fig. 3(b). The two bands interchange as kk goes from 0 to 2π2\pi and in this phase ν=1\nu=1. Note that k=0k=0 and 2π2\pi are equivalent points and in this phase E1(0)=E2(2π)E_{1}(0)=E_{2}(2\pi). Thus two energy bands form a whole circle instead of two circles. For m=2m=2, and Γ0=0.04,Γ0+=0.49,Γ11=Γ21+=0.13i,Γ21=Γ11+=0.02i,Γ12=0.58i,Γ22+=0.21i,Γ12+=0.03i,\Gamma^{0-}=0.04,\Gamma^{0+}=0.49,\Gamma^{1-}_{1}=\Gamma^{1+}_{2}=-0.13i,\Gamma^{1-}_{2}=\Gamma^{1+}_{1}=0.02i,\Gamma^{2-}_{1}=-0.58i,\Gamma^{2+}_{2}=-0.21i,\Gamma^{2+}_{1}=0.03i, and Γ22=0.09i\Gamma^{2-}_{2}=0.09i, the eigenvalues form a non-trivial braiding pattern (Fig. 3(c)) and in this phase ν=2\nu=2. The two bands encircle each other and form a structure called the Hopf link. Although in this phase each band forms a circle, they cannot be separated by any line in the complex energy plane as done in the unlink phase.

Refer to caption
Figure 3: Experimental results of energy eigenvalues for different phases. Red and blue points are experimental results for each band. Lines correspond to theoretical predictions. (a) Unlink phase with wingding number ν=0\nu=0. (b) Unknot phase with ν=1\nu=1. (c) Hopf link phase with ν=2\nu=2.

Apart from the non-trivial topology of the energy bands, the global Berry phase determined by the eigenstates was also observed. The global Berry phase is Q=02πQ=\int_{0}^{2\pi}Tr[A(k)]dk[A(k)]dk, where the non-abelian Berry connection is defined as Amn(k)=iχm(k)|k|ψn(k)A_{mn}(k)=i\braket{\chi_{m}(k)}{\partial_{k}}{\psi_{n}(k)} Q_Liang . Here |ψn(k)\ket{\psi_{n}(k)} (|χm(k)\ket{\chi_{m}(k)}) are right (left) eigenstates of an NH Hamiltonian and satisfy the biorthogonal relation χm(k)|ψn(k)=δmn\braket{\chi_{m}(k)}{\psi_{n}(k)}=\delta_{mn}. The global Berry phase QQ can identify topological invariance in our model Q_Liang . For the unlink, unknot and Hopf link phases, Qideal=0,πQ_{\rm ideal}=0,\pi and 2π2\pi, respectively. Figure 4 shows the measured eigenstates projected on the XY plane and they show the same behavior as the eigenvalues. In the unlink phase, |ψ1,2(k)\ket{\psi_{1,2}(k)} form two separated circles on the Bloch sphere, and so are the projections on the XY plane. In the unknot phase, the eigenstates form an end-to-end circle instead, just as the energy bands exchange each other. While in the Hopf link phase, each eigenstates form a closed loop, and the two loops intersect each other. After obtaining |ψ1,2(k)\ket{\psi_{1,2}(k)} from MLE, |χ1,2(k)\ket{\chi_{1,2}(k)} can be solved from the biorthogonal relation. The experimental value of QQ can be obtained with the method in Ref. ComputeQ via Q=i,nImDinQ=\sum_{i,n}\text{Im}D_{i}^{n}, where Din=lnχn(ki+1)|ψn(ki)D_{i}^{n}=\text{ln}\braket{\chi_{n}(k_{i+1})}{\psi_{n}(k_{i})} and n=1n=1 and 22 is the band index. The experimental results of QQ for the unlink, unknot and Hopf link phases are 0.00(2)π\pi, 1.03(2)π\pi and 2.00(3)π\pi, respectively, which agree well with theoretical predictions. Generally speaking, for nn bands models with eigenvalues {Ei}\{E_{i}\} and starting at k=0k=0, the ordered set (E1,E2,,En)(E_{1},E_{2},...,E_{n}) goes to (Eσ(1),Eσ(2),,Eσ(n))(E_{\sigma(1)},E_{\sigma(2)},...,E_{\sigma(n)}) as kk varies to 2π2\pi. Then we have eiQ=(1)P(σ)e^{iQ}=(-1)^{P(\sigma)}, where P(σ)P(\sigma) is the parity of the permutation. For our case, both the unlink phase and the Hopf link phase have even parity since each band returns to itself as kk goes from 0 to 2π2\pi. Thus in these two phases Q=0(Q=0~{}(mod 2π)2\pi). Since the energy bands of the unknot phase exchange each other, the parity is odd with Q=π(Q=\pi~{}(mod 2π)2\pi).

Refer to caption
Figure 4: Experimental results of the eigenstates projected on the XY plane and the global Berry phase for each phase. Dashed lines are theoretical predictions. (a) The eigenstates are separated in two regions for the unlink phase. The measured global Berry phase is Q=0.00(2)πQ=0.00(2)\pi. (b) The eigenstates form a circle for the unknot phase with Q=1.03(2)πQ=1.03(2)\pi. (c) Each eigenstate forms a closed loop, and the two loops intersect each other for the Hopf link phase. The Berry phase is measured to be Q=2.00(3)πQ=2.00(3)\pi.

In conclusion, we have experimentally investigated the knot topology in an 1D NH model based on both eigenvalues and eigenstates. The knot structures of eigenvalues, including the unlink, unknot and Hopf link phases, were successfully observed, which manifest the B(2) braid group behavior. The global Berry phase was measured via high fidelity eigenstates, which serves as the knot invariant to identify the parity of band braiding. Our work makes NV center a desirable platform for investigating important non-Hermitian topology. The universality of our dilation method for arbitrary dimensional cases and the ground-state three-level structure of the NV center make it possible to explore the knot topology of three-band models. For 1D models with three bands, the knotted topological phases are described by the conjugate classes of B(3)B(3) PRL_Haiping ; NatPhys_Bouhon , which host richer topological behaviors since B(N)B(N) is not commutative when N>2N>2. By introducing more momentum space parameters such as k=(kx,ky,kz)\vec{k}=(k_{x},k_{y},k_{z}), our platform can be utilized to investigate the knot topology of higher dimensional NH models PRB_Fan ; PRB_ZhiLi .

This work was supported by the National Key R&D Program of China (Grant Nos. 2018YFA0306600 and 2016YFB0501603), the National Natural Science Foundation of China (Grant No. 12174373), the Chinese Academy of Sciences (Grant Nos. XDC07000000 and GJJSTD20200001), Innovation Program for Quantum Science and Technology (Grant No. 2021ZD0302200), Anhui Initiative in Quantum Information Technologies (Grant No. AHY050000) and Hefei Comprehensive National Science Center. X. R. thanks the Youth Innovation Promotion Association of Chinese Academy of Sciences for their support. Ya Wang and Yang Wu thanks the Fundamental Research Funds for the Central Universities for their support.

Yang Wu, Yunhan Wang and Xiangyu Ye contributed equally to this work.

I Appendix A: Experimental Setup

The experimental setup is shown in Fig.5. The diamond was mounted on a confocal setup, and the static magnetic field of 506 G was provided by a permanent magnet along the NV symmetry axis. The initialization and the readout of the NV center spin were realized with a 532 nm green laser controlled by an acousto-optic modulator (AOM) (ISOMET, AOMO 3200-121). The laser beam traveled twice through the acousto-optic modulator before going through an oil objective (Olympus, PLAPON 60*O, NA 1.42) and then focusing on the NV center. The phonon sideband fluorescence (wavelength, 650-800nm) went through the same oil objective and was collected by an avalanche photodiode (Perkin Elmer, SPCM-AQRH-14) with a counter card (NI, PCIe-6612).

The radio-frequency (RF) pulses were generated by an arbitrary waveform generator (AWG) (CIQTEK AWG4100) and were amplified by a power amplifier (Mini-Circuits, LZY-22+). The microwave (MW) pulses were generated by the same arbitrary waveform generator. The bandwidth of the arbitrary waveform generator is 0-330 MHz, which is far from the resonant frequency (about 1.47 GHz). Thus, the pulses were mixed with continuous 1.6-GHz output from a wave source (RIGOL, DSG3065B) using an IQ modulator (Marki, IQ1545LMP). Then the pulses passed a PIN (Mini-Circuits, ZASWA-2-50DRA+) and an amplifier (Mini-Circuits, ZHL-15W-422-S+). Finally both the MW and RF pulses were fed by the same coplanar waveguide after passing the diplexer (Marki, DPX-0R5). The waveforms of MW and RF pulses were prepared in advance and were downloaded to the AWG. An arbitrary sequence generator (CIQTEK ASG8100) was utilized to control the timing in experiment by the trigger signals designed in advance.

Refer to caption
Figure 5: The hardware setup in our experiment. Both MW and RF pulses were generated by the AWG and an IQ modulator was utilized to adjust the frequency of MW pulses to the transition frequency we needed. The pulses were fed by the same coplanar waveguide after being amplified by amplifiers. The 532 nm laser was used to polarize and readout the spin state of the NV center. The on/off of the laser beam was controlled by an AOM. The fluorescence (650-800 nm) from the NV center passed the objective and was collected by an APD. The timing for our experiment was controlled by the arbitrary sequence generator (ASG).

II Appendix B: Universal Dilation Method

We use the universal dilation method to construct the NH Hamiltonian Science_YW . Intuitively, by introducing an ancilla system and tuning their interaction in a time-dependent way, the target system obeys the evolution governed by the NH Hamiltonian HsH_{s}. The detail can be found in Ref. Science_YW . For an NH Hamiltonian HsH_{s}, the dilated Hamiltonian Hs,aH_{s,a} takes the form:

Hs,a=Λ(t)I+Γ(t)σz,H_{s,a}=\Lambda(t)\otimes I+\Gamma(t)\otimes\sigma_{z}, (4)

where II is the identity matrix and σz\sigma_{z} is the Pauli operator on the ancilla. Λ(t)\Lambda(t) and Γ(t)\Gamma(t) are operators on the system written as: Γ(t)={Hs(t)+[iddtη(t)+η(t)Hs(t)]η(t)}M1(t)\Gamma(t)=\{H_{s}(t)+[i\frac{d}{dt}\eta(t)+\eta(t)H_{s}(t)]\eta(t)\}M^{-1}(t) and Λ(t)=i[Hs(t)η(t)η(t)Hs(t)iddtη(t)]M1(t)\Lambda(t)=i[H_{s}(t)\eta(t)-\eta(t)H_{s}(t)-i\frac{d}{dt}\eta(t)]M^{-1}(t). Here M(t)=η(t)η(t)+IM(t)=\eta(t)^{\dagger}\eta(t)+I and M(t)M(t) satisfies the following equation:

iddtM(t)=Hs(t)M(t)M(t)Hs(t).i\frac{d}{dt}M(t)=H_{s}^{\dagger}(t)M(t)-M(t)H_{s}(t). (5)

Now we focus on the realization of Hs,aH_{s,a} in the NV center. The ground state of the NV center can be described by the Hamiltonian

HNV=2π(DSz2+ωeSz+QIz2+ωnIz+ASzIz),H_{NV}=2\pi(DS_{z}^{2}+\omega_{e}S_{z}+QI_{z}^{2}+\omega_{n}I_{z}+AS_{z}I_{z}), (6)

where Sz(Iz)S_{z}\ (I_{z}) is the spin-1 operator of the electron (nuclear) spin, D=2.87D=2.87 GHz is the zero field splitting for the electron, ωe(ωn)\omega_{e}\ (\omega_{n}) is the electron (nuclear) Zeeman term induced by the magnetic field applied along the NV symmetry axis, Q=4.95Q=-4.95 MHz is the nuclear quadrupolar interaction, and A=2.16A=-2.16 MHz is the hyperfine interaction. We construct the dilated Hamiltonian HtotH_{\rm tot} in the subspace spanned by the states |mS,mI=|0,1,|0,0,|1,1\ket{m_{S},m_{I}}=\ket{0,1},\ket{0,0},\ket{-1,1}, and |1,0\ket{-1,0}. The Hamiltonian in this subspace can be simplified to

H0=π[(DωeA2)σzI+(Q+ωnA2)Iσz+A2σzσz].H_{0}=\pi[-(D-\omega_{e}-\frac{A}{2})\sigma_{z}\otimes I+(Q+\omega_{n}-\frac{A}{2})I\otimes\sigma_{z}+\frac{A}{2}\sigma_{z}\otimes\sigma_{z}]. (7)

To facilitate the construction of Hs,a(t)H_{s,a}(t), we further decompose Hs,a(t)H_{s,a}(t) as:

Hs,a=B1II+A1σxI+B2σyI+B3σzI+A2Iσz+B4σxσz+A3σyσz+A4σzσz,\begin{split}H_{s,a}&=B_{1}I\otimes I+A_{1}\sigma_{x}\otimes I+B_{2}\sigma_{y}\otimes I+B_{3}\sigma_{z}\otimes I\\ &+A_{2}I\otimes\sigma_{z}+B_{4}\sigma_{x}\otimes\sigma_{z}+A_{3}\sigma_{y}\otimes\sigma_{z}+A_{4}\sigma_{z}\otimes\sigma_{z},\end{split} (8)

where AiA_{i} and BiB_{i} are time-dependent coefficients. We apply selective MW pulses to cast the control Hamiltonian:

Hc(t)=2πΩ1(t)cos[0tω1(τ)𝑑τ+ϕ1(t)]σx|1n1|+2πΩ2(t)cos[0tω2(τ)𝑑τ+ϕ2(t)]σx|0n0|.\begin{split}H_{c}(t)&=2\pi\Omega_{1}(t)\cos[\int_{0}^{t}\omega_{1}(\tau)d\tau+\phi_{1}(t)]\sigma_{x}\otimes\ket{1}_{n}\bra{1}\\ &+2\pi\Omega_{2}(t)\cos[\int_{0}^{t}\omega_{2}(\tau)d\tau+\phi_{2}(t)]\sigma_{x}\otimes\ket{0}_{n}\bra{0}.\end{split} (9)

In order to realize Hs,aH_{s,a}, we choose the rotating frame with the following form:

Urot=ei0tH0B1(τ)IIB3(τ)σzIA2(τ)IσzA4(τ)σzσzdτ.U_{rot}=e^{i\int_{0}^{t}H_{0}-B_{1}(\tau)I\otimes I-B_{3}(\tau)\sigma_{z}\otimes I-A_{2}(\tau)I\otimes\sigma_{z}-A_{4}(\tau)\sigma_{z}\otimes\sigma_{z}d\tau}. (10)

Thus the Hamiltonian in the rotating frame can be written as

Htot=Urot(H0+Hc)UrotiUrotdUrotdt=B1II+B3σzI+A2Iσz+A4σzσz+2πΩ1cos(ϕ1+0tω1)cos(0tω~1+2B3+2A4)σx|1n1|+2πΩ1cos(ϕ1+0tω1)sin(0tω~1+2B3+2A4)σy|1n1|+2πΩ2cos(ϕ2+0tω2)cos(0tω~2+2B32A4)σx|0n0|+2πΩ2cos(ϕ2+0tω2)sin(0tω~2+2B32A4)σy|0n0|.\begin{split}&H_{tot}=U_{rot}(H_{0}+H_{c})U_{rot}^{\dagger}-iU_{rot}\frac{dU_{rot}^{\dagger}}{dt}\\ &=B_{1}I\otimes I+B_{3}\sigma_{z}\otimes I+A_{2}I\otimes\sigma_{z}+A_{4}\sigma_{z}\otimes\sigma_{z}\\ &+2\pi\Omega_{1}\cos(\phi_{1}+\int_{0}^{t}\omega_{1})\cos(\int_{0}^{t}\tilde{\omega}_{1}+2B_{3}+2A_{4})\sigma_{x}\otimes\ket{1}_{n}\bra{1}\\ &+2\pi\Omega_{1}\cos(\phi_{1}+\int_{0}^{t}\omega_{1})\sin(\int_{0}^{t}\tilde{\omega}_{1}+2B_{3}+2A_{4})\sigma_{y}\otimes\ket{1}_{n}\bra{1}\\ &+2\pi\Omega_{2}\cos(\phi_{2}+\int_{0}^{t}\omega_{2})\cos(\int_{0}^{t}\tilde{\omega}_{2}+2B_{3}-2A_{4})\sigma_{x}\otimes\ket{0}_{n}\bra{0}\\ &+2\pi\Omega_{2}\cos(\phi_{2}+\int_{0}^{t}\omega_{2})\sin(\int_{0}^{t}\tilde{\omega}_{2}+2B_{3}-2A_{4})\sigma_{y}\otimes\ket{0}_{n}\bra{0}.\end{split} (11)

We have omitted the tt dependence and the integral variable for simplicity, and ω~1\tilde{\omega}_{1} (ω~2)(\tilde{\omega}_{2}) is the transition frequency of between |0,1\ket{0,1} and |1,1\ket{-1,1} (|0,0\ket{0,0} and |1,0\ket{-1,0}). In order to reduce HtotH_{tot} to Hs,aH_{s,a} under the rotating-wave approximation, the amplitudes, frequencies and phases should satisfy ω1=ω~1+2B3+2A4,ω2=ω~2+2B32A4\omega_{1}=\tilde{\omega}_{1}+2B_{3}+2A_{4},\omega_{2}=\tilde{\omega}_{2}+2B_{3}-2A_{4} and

{π2(Ω1cosϕ1+Ω2cosϕ2)=A1,π2(Ω1cosϕ1Ω2cosϕ2)=B4,π2(Ω1sinϕ1Ω2sinϕ2)=B2,π2(Ω1sinϕ1+Ω2sinϕ2)=A3.\left\{\begin{aligned} &\frac{\pi}{2}(\Omega_{1}\cos\phi_{1}+\Omega_{2}\cos\phi_{2})=A_{1},\\ &\frac{\pi}{2}(\Omega_{1}\cos\phi_{1}-\Omega_{2}\cos\phi_{2})=B_{4},\\ &\frac{\pi}{2}(-\Omega_{1}\sin\phi_{1}-\Omega_{2}\sin\phi_{2})=B_{2},\\ &\frac{\pi}{2}(-\Omega_{1}\sin\phi_{1}+\Omega_{2}\sin\phi_{2})=A_{3}.\end{aligned}\right. (12)

The solution reads

{Ω1=(A1+B4)2+(B2+A3)2π,Ω2=(A1B4)2+(B2+A3)2π,ϕ1=atan2(B2A3,A1+B4),ϕ2=atan2(A3B2,A1B4).\left\{\begin{aligned} \Omega_{1}&=\frac{\sqrt{(A_{1}+B_{4})^{2}+(B_{2}+A_{3})^{2}}}{\pi},\\ \Omega_{2}&=\frac{\sqrt{(A_{1}-B_{4})^{2}+(-B_{2}+A_{3})^{2}}}{\pi},\\ \phi_{1}&=\text{atan2}(-B_{2}-A_{3},A_{1}+B_{4}),\\ \phi_{2}&=\text{atan2}(A_{3}-B_{2},A_{1}-B_{4}).\end{aligned}\right. (13)

As an example, the result for the Hopf link phase with k=1.65πk=1.65\pi is shown in Fig. 6. The target NH Hamiltonian is

H(2)(k)=[0Γ0Γ0+0]+\displaystyle H^{(2)}(k)=\left[\begin{array}[]{ccc}0&\Gamma^{0-}\\ \Gamma^{0+}&0\end{array}\right]+ (14)
n=12[0Γ1neink+Γ2n+einkΓ1n+eink+Γ2neink0],\displaystyle\sum_{n=1}^{2}\left[\begin{array}[]{ccc}0&\Gamma_{1}^{n-}e^{ink}+\Gamma_{2}^{n+}e^{-ink}\\ \Gamma_{1}^{n+}e^{-ink}+\Gamma_{2}^{n-}e^{ink}&0\end{array}\right],

where Γ0=0.04,Γ0+=0.49,Γ11=Γ21+=0.13i\Gamma^{0-}=0.04,\Gamma^{0+}=0.49,\Gamma_{1}^{1-}=\Gamma_{2}^{1+}=-0.13i, Γ21=Γ11+=0.02i,Γ12=0.58i,Γ22+=0.21i\Gamma_{2}^{1-}=\Gamma_{1}^{1+}=0.02i,\Gamma_{1}^{2-}=-0.58i,\Gamma_{2}^{2+}=-0.21i, Γ12+=0.03i,Γ22=0.09i\Gamma_{1}^{2+}=0.03i,\Gamma_{2}^{2-}=0.09i, and k=1.65π.k=1.65\pi. Since the eigenvalues of the Hamiltonian are complex, the state evolution shows a damping behavior and finally reaches one of the eigenstate of the NH Hamiltonian HsH_{s}. The dilation procedure gives Ωi(t)\Omega_{i}(t) and ϕi(t)\phi_{i}(t) to realize the dilated Hamiltonian HtotH_{tot}, as shown in Fig. 6. For an intuitive interpretation, both Ωi(t)\Omega_{i}(t) and ϕi(t)\phi_{i}(t) first show a decay behavior and then almost remain unchanged. When Ωi(t)\Omega_{i}(t) and ϕi(t)\phi_{i}(t) are almost equal to their steady value, they effectively act as a rotation along a specific axis determined by the eigenstate of the NH Hamiltonian in the target subspace. The beginning decay part of the control Hamiltonian exactly drives the initial state to be parallel with this rotation axis. The final state thus remains unchanged (in the target subspace) though Ωi\Omega_{i} is not zero. The measured population evolution agrees well with this picture and the theoretical predictions.

Refer to caption
Figure 6: The time-dependent amplitudes and phases of the control pulses and the corresponding experimental result of the population evolution. The results correspond to the Hopf link phase with k=1.65πk=1.65\pi.
Refer to caption
Figure 7: Simulation results of population evolution in the σx\sigma_{x} (a,c,e) basis and the σy\sigma_{y} (b,d,f) basis. (a-d) NV centers in diamond with 12C natural abundance and different λ\lambda’s. (e,f) Our parameters and NV center in 12C purified diamond. The parameters are as follows: (a,b) λ=2π×850\lambda=2\pi\times 850 kHz, T2=1.5μT_{2}^{\star}=1.5\ \upmus; (c,d) λ=2π×2550\lambda=2\pi\times 2550 kHz, T2=1.5μT_{2}^{\star}=1.5\ \upmus; and (e,f) A=2.16A=-2.16 MHz, T2=78μT_{2}^{\star}=78\ \upmus, and λ=2π×85\lambda=2\pi\times 85 kHz.

III Appendix C: Effects of Dephasing and Crosstalk

The evolution under the NH Hamiltonian is mainly affected by the dephasing and the crosstalk. Equation 11 shows the ideal case when the two MW pulses are selective. However, in practice the operators in the control Hamiltonian HcH_{c} have the form σxI\sigma_{x}\otimes I instead of σx|1n1|\sigma_{x}\otimes\ket{1}_{n}\bra{1}, and the decoherence inevitably undermines the evolution. As mentioned in the main text, one way is to use a strongly coupled 13C nuclear spin. For a natural abundance diamond, the dephasing time T2T_{2}^{\star} ranges from 1μ1\ \upmus to 3μ3\ \upmus. We take T2=1.5μT_{2}^{\star}=1.5\ \upmus as a typical value. The coupling strength ACA_{C} between the 13C nuclear spin and the electron spin is on the order of 10 MHz generally and we take |AC|=15|A_{C}|=15 MHz. The simulation results of the evolution under the σx\sigma_{x} basis for the Hopf link phase with k=0.85πk=0.85\pi and different λ\lambda’s are shown in Fig. 7. If one uses weak driving to suppress the crosstalk, the evolution time is 1.5μ1.5\ \upmus, which is comparable to the coherence time. The decoherence drastically destroys the coherent evolution. The population evolution under σx,y\sigma_{x,y} bases is shown in Figs. 7(a) and 7 (b), which greatly deviates from the ideal case. We use c=σx2+σy2c=\braket{\sigma_{x}}^{2}+\braket{\sigma_{y}}^{2} to characterize the coherence. In this parameter configuration, for the final state we have c=0.65c=0.65 and for the ideal case we have c=0.97c=0.97. Only about 67% coherence is left. One may try to set a large value of λ\lambda under which the time needed to reach the steady state is within the coherence time. When λ/2π\lambda/2\pi varies from 850850 kHz to 25502550 kHz, λ/|AC|\lambda/|A_{C}| increases from 0.057 to 0.17 and the evolution time needed decreases from 1.7μ1.7\ \upmus to 0.6μ0.6\ \upmus. The pulses are no longer selective and the evolution is significantly affected by the crosstalk. As can be seen from Fig. 7(c) and 7(d), the system can barely reach a steady state.

Another way is to use samples in which the NV center has a long coherence time. For our sample, fitting of the luminescence in the Ramsey experiment shows that T2=78(7)μT_{2}^{\star}=78(7)\ \upmus (see main text). Fig. 7(e) and 7(f) show the simulation results for our parameters where we take T2=78μT_{2}^{\star}=78\ \upmus. Here A=2.16A=-2.16 MHz is a typical value for the coupling strength between the 14N nuclear spin and the electron spin. For this parameter configuration, λ/|A|=0.039\lambda/|A|=0.039. The deviation caused by decoherence and crosstalk is about 0.03 and these effects on the fidelity between the final states in experiments and the ideal eigenstates can be ignored.

IV Appendix D: Experimental Acquisition of the Eigenstates

Refer to caption
Figure 8: Measurement sequences to reconstruct the state. Counts C1C9C_{1}\sim C_{9} are measured counts for each sequence.

For the final state of the evolution, nine different measurement sequences are applied to reconstruct the state (Fig. 8). Here Re(π)R^{e}(\pi) is the π\pi pulse on the electron spin in the |1n\ket{1}_{n} subspace and Rn(π)R^{n}(\pi) is the π\pi pulse on the nuclear spin in the |0e\ket{0}_{e} subspace. Re(π)R^{e}(\pi) and Rn(π)R^{n}(\pi) correspond to the transition MW1 and RF2, respectively, in Fig.2(d) of the main text. The π/2\pi/2 pulses rotate the electron spin in both spaces. We use normalization sequences to obtain the PL rate for each of the four levels PRL_Wenquan . The contribution for the counts of each level is the PL rate multiplied by the corresponding population. The count of each measurement sequence equals the summation of the contribution over each level. By solving the equations that relate the populations of each level under different bases and counts for each measurement sequence, we can obtain the expectation values σx,y,z\braket{\sigma_{x,y,z}} of the final state.

From σx,y,z\braket{\sigma_{x,y,z}} of the final state, we can directly reconstruct ρ\rho by ρ=(I+σσ)/2\rho=(I+\braket{\vec{\sigma}}\vec{\sigma})/2, but the direct result may give a mixed state or even an unphysical state. Thus the maximum likelihood estimation has been utilized to obtain the pure states from the experimental results. We parameterize the pure state as (α|0e+βeiγ|1e)|1n+(δ|0e+ϵeiζ|1e)|0n(\alpha\ket{0}_{e}+\beta e^{i\gamma}\ket{-1}_{e})\ket{1}_{n}+(\delta\ket{0}_{e}+\epsilon e^{i\zeta}\ket{-1}_{e})\ket{0}_{n}, where all parameters are real and satisfy the normalization condition. Note that the measurement result of each sequence is determined by the population on each level. Since Rn(π)R^{n}(\pi) and Re(π)R^{e}(\pi) reverse the population of the corresponding levels, the phase difference between the two subspace does not manifest in the measurement result. Here we fix the coefficients of |0e\ket{0}_{e} to be real to eliminate the irrelevant phase freedom, since we only need the results in the |1n\ket{1}_{n} subspace. Then the expectation values for the nine counts can be obtained from the PL rates p1,2,3,4p_{1,2,3,4} and the parameters of the pure state, where we label the levels |mS,mI=|0,1,|1,1,|0,0\ket{m_{S},m_{I}}=\ket{0,1},\ket{-1,1},\ket{0,0}, and |1,0\ket{-1,0} as 1,2,3,1,2,3, and 44 for simplicity. For example, the expectation values for the counts of the first and second sequences read C~1=α2p1+β2p2+δ2p3+ϵ2p4\tilde{C}_{1}=\alpha^{2}p_{1}+\beta^{2}p_{2}+\delta^{2}p_{3}+\epsilon^{2}p_{4} and C~2=β2p1+α2p2+δ2p3+ϵ2p4\tilde{C}_{2}=\beta^{2}p_{1}+\alpha^{2}p_{2}+\delta^{2}p_{3}+\epsilon^{2}p_{4}. Then the loss function is chosen to be

L(α,β,γ,δ,ϵ,ζ)=i=19(CiC~i)2,L(\alpha,\beta,\gamma,\delta,\epsilon,\zeta)=\sum_{i=1}^{9}(C_{i}-\tilde{C}_{i})^{2}, (15)

where CiC_{i} are the measured counts for each sequence. Optimize these parameters to minimize LL and we obtain the pure state αmin|0e+βmineiγmin|1e\alpha_{\text{min}}\ket{0}_{e}+\beta_{\text{min}}e^{i\gamma_{\text{min}}}\ket{-1}_{e} up to a normalization factor. Here the subscript means the parameters that minimize LL. The experimentally obtained fidelities of all the eigenstates exceed 0.97. Based on the model parameters given in the main text, we show the results of some eigenstates ψ1,2\psi_{1,2} as examples in Fig.r̃efSFig3. The corresponding fidelities are as follows: Hopf link, 1.00(7) and 1.00(6) for ψ1\psi_{1} and ψ2\psi_{2} (same below) at k=0.1πk=0.1\pi; unknot, 1.00(2) and 1.00(3) at k=πk=\pi and unlink, 1.00(3) and 1.00(3) at k=0.6πk=0.6\pi.

Refer to caption
Figure 9: Examples for measured states in each phase. (a,b) The Hopf link phase with k=0.1πk=0.1\pi. Panel (a) shows the real and imaginary parts for the one eigenstate, and panel (b) shows the real and imaginary parts of the other eigenstate of the same Hamiltonian. The arrangement of other sub-figures is similar. (c,d) The unknot phase with k=πk=\pi. (e,f) The unlink phase with k=0.6πk=0.6\pi.

References

  • (1) E. J. Bergholtz, J. C. Budich, and F. K. Kunst, Exceptional topology of non-Hermitian systems, Rev. Mod. Phys. 93, 015005 (2021).
  • (2) S. Yao, and Z. Wang, Edge states and topological invariants of non-Hermitian systems, Phys. Rev. Lett. 121, 086803 (2018).
  • (3) K. Kawabata, K. Shiozaki, M. Ueda, and M. Sato, Symmetry and topology in non-Hermitian physics, Phys. Rev. X 9, 041015 (2019).
  • (4) Z. Gong, Y. Ashida, K. Kawabata, K. Takasan, S. Higashikawa, and M. Ueda, Topological phases of non-Hermitian systems, Phys. Rev. X 8, 031079 (2018).
  • (5) T. E. Lee, Anomalous edge state in a non-Hermitian lattice, Phys. Rev. Lett. 116, 133903 (2016).
  • (6) H. Zhou, C. Peng, Y. Yoon, C. W. Hsu, K. A. Nelson, L. Fu, J. D. Joannopoulos, M. Soljacˇ\check{\text{c}}ic´\acute{\text{c}}, and B. Zhen, Observation of bulk Fermi arc and polarization half charge from paired exceptional points, Science 359, 1009-1012 (2018).
  • (7) J. Carlstro¨\ddot{\text{o}}m, and E. J. Bergholtz, Exceptional links and twisted Fermi ribbons in non-Hermitian systems, Phys. Rev. A 98, 042114 (2018).
  • (8) B. Peng, Ş. K. O¨\ddot{\text{O}}zdemir, F. Lei, F. Monifi, M. Gianfreda, G. L. Long, S. Fan, F. Nori, C. M. Bender, and L. Yang, Parity-time-symmetric whispering-gallery microcavities, Nature Physics 10, 394-398 (2014).
  • (9) L. Feng, M. Ayache, J. Huang, Y.-L. Xu, M.-H. Lu, Y.-F. Chen, Y. Fainman, and A. Scherer, Nonreciprocal light propagation in a silicon photonic circuit, Science 333, 729-733 (2011).
  • (10) L. Feng, Z. J. Wong, R.-M. Ma, Y. Wang, and X. Zhang, Single-mode laser by parity-time symmetry breaking, Science 346, 972-975 (2014).
  • (11) H. Hodaei, M.-A. Miri, M. Heinrich, D. N. Christodoulides, and M. Khajavikhan, Parity-time-symmetric microring lasers, Science 346, 975-978 (2014).
  • (12) R. El-Ganainy, K. G. Makris, M. Khajavikhan, Z. H. Musslimani, S. Rotter, and D. N. Christodoulides, Non-Hermitian physics and PT symmetry, Nature Physics 14, 11-19 (2018).
  • (13) K. G. Makris, R. El-Ganainy, D. N. Christodoulides, and Z. H. Musslimani, Beam Dynamics in PT Symmetric Optical Lattices, Phys. Rev. Lett. 100, 103904 (2008).
  • (14) H. Jing, S. K. O¨\ddot{\text{O}}zdemir, X.-Y. Lu¨\ddot{\text{u}}, J. Zheng, L. Yang, and F. Nori, PT-Symmetric Phonon Laser, Phys. Rev. Lett. 113, 053604 (2014).
  • (15) J. M. Zeuner, M. C. Rechtsman, Y. Plotnik, Y. Lumer, S. Nolte, M. S. Rudner, M. Segev, and A. Szameit, Observation of a topological transition in the bulk of a non-Hermitian system, Phys. Rev. Lett. 115, 040402 (2015).
  • (16) L. Feng, R. El-Ganainy, and L. Ge, Non-Hermitian photonics based on parity-time symmetry, Nature Photonics 11, 752-762 (2017).
  • (17) A. Regensburger, C. Bersch, M.-A. Miri, G. Onishchukov, D. N. Christodoulides, and U. Peschel, Parity-time synthetic photonic lattices, Nature 488, 167-171 (2012).
  • (18) B. Peng, Ş. K. O¨\ddot{\text{O}}zdemir, S. Rotter, H. Yilmaz, M. Liertzer, F. Monifi, C. M. Bender, F. Nori, and L. Yang, Loss-induced suppression and revival of lasing, Science 346, 328-332 (2014).
  • (19) R. Fleury, D. Sounas, and A. Alu`\grave{\text{u}}, An invisible acoustic sensor based on parity-time symmetry, Nature Commun. 6, 5905 (2015).
  • (20) T. Liu, X. Zhu, F. Chen, S. Liang, and J. Zhu, Unidirectional Wave Vector Manipulation in Two-Dimensional Space with an All Passive Acoustic Parity-Time-Symmetric Metamaterials Crystal, Phys. Rev. Lett. 120, 124502 (2018).
  • (21) Z. Zhang, M. R. Lo´\acute{\text{o}}pez, Y. Cheng, X. Liu, and J. Christensen, Non-Hermitian Sonic Second-Order Topological Insulator, Phys. Rev. Lett. 122, 195501 (2019).
  • (22) I. Rotter, A non-Hermitian Hamilton operator and the physics of open quantum systems, J. Phys. A 42, 153001 (2009).
  • (23) B. Zhen, C. W. Hsu, Y. Igarashi, L. Lu, I. Kaminer, A. Pick, S.-L. Chua, J. D. Joannapoulos, and M. Soljacˇ\check{\text{c}}ic´\acute{\text{c}}, Spawning rings of exceptional points out of Dirac cones, Nature 525, 354-358 (2015).
  • (24) H. J. Carmichael, Quantum trajectory theory for cascaded open systems, Phys. Rev. Lett. 70, 2273 (1993).
  • (25) S. Diehl, E. Rico, M. A. Baranov, and P. Zoller, Topology by dissipation in atomic quantum wires, Nature Physics 7, 971-977 (2011).
  • (26) V. Kozii, and L. Fu, Non-hermitian topological theory of finite-lifetime quasiparticles: Prediction of bulk Fermi arc due to exceptional point, arXiv:1708.05841.
  • (27) T. Yoshida, R. Peters, and N. Kawakami, Non-Hermitian perspective of the band structure in heavy-fermion systems, Phys. Rev. B 98, 035141 (2018).
  • (28) H. Shen, and L. Fu, Quantum oscillation from in-gap states and a non-Hermitian Landau level problem, Phys. Rev. Lett. 121, 026403 (2018).
  • (29) M. Papaj, H. Isobe, and L. Fu, Nodal arc of disordered Dirac fermions and non-Hermitian band theory, Phys. Rev. B 99, 201107(R) (2019).
  • (30) K. Zhang, Z. Yang, and C. Fang, Correspondence between winding numbers and skin modes in non-Hermitian systems, Phys. Rev. Lett. 125, 126402 (2020).
  • (31) D. S. Borgnia, A. J. Kruchkov, and R.-J. Slager Non-Hermitian boundary modes and topology, Phys. Rev. Lett. 124, 056802 (2020).
  • (32) N. Okuma, K. Kawabata, K. Shiozaki, and M. Sato, Topological origin of non-Hermitian skin effects, Phys. Rev. Lett. 124, 086801 (2020).
  • (33) C. C. Wojcik, X.-Q. Sun, T. Bzdusˇ\check{\text{s}}ek, and S. Fan, Homotopy characterization of non-Hermitian Hamiltonians, Phys. Rev. B 101, 205417 (2020).
  • (34) H. Hu, and E. Zhao, Knots and non-Hermitian Bloch bands, Phys. Rev. Lett. 126, 010401 (2021).
  • (35) Z. Li, and R. S. K. Mong, Homotopical characterization of non-Hermitian band structures, Phys. Rev. B 103, 155129 (2021).
  • (36) K. Wang, A. Dutt, C. C. Wojcik, and S. Fan, Topological complex-energy braiding of non-Hermitian bands, Nature 598, 59-64 (2021).
  • (37) Y. S. S. Patil, J. Ho¨\ddot{\text{o}}ller, P. A. Henry, C. Guria, Y. Zhang, L. Jiang, N. Kralj, N. Read, and J. G. E. Harris, Measuring the knot of non-Hermitian degeneracies and non-commuting braids, Nature 607, 271-275 (2022).
  • (38) M.-M. Cao, K. Li, W.-D. Zhao, W.-X. Guo, B.-X. Qi, X.-Y. Chang, Z.-C. Zhou, Y. Xu, and L.-M. Duan, Probing complex-energy topology via non-Hermitian absorption spectroscopy in a trapped ion simulator, Phys. Rev. Lett. 130, 163001 (2023).
  • (39) Y. Wu, W. Liu, J. Geng, X. Song, X. Ye, C. Duan, X. Rong, and J. Du, Observation of parity-time symmetry breaking in a single-spin system, Science 364, 878-880 (2019).
  • (40) W. Zhang, X. Ouyang, X. Huang, X. Wang, H. Zhang, Y. Yu, X. Chang, Y. Liu, D.-L. Deng, and L.-M. Duan, Observation of non-Hermitian topology with non-unitary dynamics of solid-state spins, Phys. Rev. Lett. 127, 090501 (2021).
  • (41) U. Gu¨\ddot{\text{u}}nther, and B. F. Samsonov, Naimark-dilated PT-symmetric brachistochrone, Phys. Rev. Lett. 101, 230404 (2008).
  • (42) K. Kawabata, Y. Ashida, and M. Ueda, Information retrieval and criticality in parity-time-symmetric systems, Phys. Rev. Lett. 119, 190401 (2017).
  • (43) V. Jacques, P. Neumann, J. Beck, M. Markham, D. Twitchen, J. Meijer, F. Kaiser, G. Balasubramanian, F. Jelezko, and J. Wrachtrup, Dynamic polarization of single nuclear spins by optical pumping of nitrogen-vacancy color centers in diamond at room temperature, Phys. Rev. Lett. 102, 057403 (2009).
  • (44) S.-D. Liang, and G.-Y. Huang, Topological invariance and global Berry phase in non-Hermitian systems, Phys. Rev. A 87, 012118 (2013).
  • (45) T. Fukui, Y. Hatsugai, and H. Suzuki, Chern numbers in discretized brillouin zone: Efficient method of computing (spin) hall conductances, J. Phys. Soc. Jpn. 74, 1674-1677 (2005).
  • (46) A. Bouhon, Q. Wu, R.-J. Slager, H. Weng, O. V. Yazyev, and T. Bzdusek, Non-Abelian reciprocal braiding of Weyl points and its manifestation in ZrTe, Nature Physics 16, 1137–1143 (2020)
  • (47) W. Liu, Y. Wu, C. Duan, X. Rong, and J. Du, Dynamically encircling an exceptional point in a real quantum system. Phys. Rev. Lett. 126, 170506 (2021).