Observing the exponential growth of the eigenmodes in the absence of coalescence for a non-Hermitian circuit with an unavoidable inductor dissipation Jiaming Zhang, Wenxuan Song, Hanhao Li, Zhiye Kuang, Zhihua Xiong, Longwen Zhou, Jie Liu, Wen-Lei Zhao Abstract We investigate, both experimentally and theoretically, the eigenmodes of an electronic circuit in which gain and loss R ​ L ​ C RLC resonators are coupled through a capacitor. Due to the unavoidable magnetic loss in the inductors, we find that the eigenmode coalescence no longer emerges in contrast to the conventional non-Hermitian systems with the spontaneous 𝒫𝒯 \cal{PT} -symmetry breaking. In particular, we find a transition from the exponential decay to exponential growth in the amplitude of the periodic voltage oscillations of the resonators. The transition occurs near the exceptional points of the non-Hermitian circuit without considering the dissipations in inductors. We introduce a small resistor of two orders of magnitude smaller than that of the R ​ L ​ C RLC resonators to mimic the energy dissipation in inductors and numerically solve the equivalent non-Hermitian Schrödinger equation. The numerical results can well reproduce experimental observations. Our above findings unambiguously indicate that the exponential growth behavior beyond the exceptional points is robust against some unavoidable dissipative perturbations. I Introduction The fundamental problems arising from exceptional points (EPs) of non-Hermitian systems have received significant attention, both theoretically and experimentally, in different fields of physics, such as quantum metrology [ 1 , 2 ] , optics [ 3 , 4 , 5 ] , as well as condensed matter of physics [ 6 , 7 , 8 ] . EPs act as branch points in complex plane, where encircling them in parameter space even induces a quantized accumulation of the Berry phase [ 9 , 10 , 11 ] , uncovering the non-Hermitian topology [ 12 , 13 , 14 ] . These topologically nontrivial singularities have profound implications for quantum transport in open systems [ 15 , 16 , 17 , 18 ] , leading, for instance, to the emergence of geometrically dependent skin effects in phononic crystals [ 19 ] , self-acceleration in dissipative quantum walks [ 20 ] , and delocalization-localization transitions in quasicrystals [ 21 , 22 , 23 , 24 ] , to name a few. EPs have been exploited to manipulate the unidirectional reflectionless light propagation [ 25 ] and perfect radiation in photonic devices [ 26 ] . Near the EPs, the square-root relationship between eigenenergy level splitting and system parameters amplifies significantly small perturbations, thereby dramatically enhancing the sensitivity of quantum sensors in various settings, including optical systems [ 27 ] ,optomechanical sensors [ 28 ] , and electronic circuits [ 29 ] . Recently, electronic circuits, offering precise control over key parameters, have emerged as powerful platforms for quantum simulations [ 30 , 31 , 32 , 33 , 34 ] . The high controllability of electronic components enables the effective emulation of diverse quantum systems and the exploration of exotic phenomena that are difficult to study in natural systems due to practical limitations. The coupled RLC resonators, with balanced gain and loss, demonstrate the coalescence of eigenvalues of \cal{PT} -symmetric systems at EPs [ 35 , 36 ] . By carefully designing RLC resonator arrays with specific connectivity patterns, one can implement paradigm models in fundamental physics, including the quantum walk [ 37 ] , the Su-Schrieffer-Heeger lattice [ 38 ] , and the Aubry-André model [ 39 ] . This paves the way for investigating novel physics like transient PT symmetry [ 40 ] , topological insulating phases [ 41 ] , and non-Abelian Inverse Anderson transitions [ 42 ] . Intrinsically, the versatility of electrical circuits, incorporating a tunable capacitor or inductor in the time domain, allows for the realization of Floquet \cal{PT} -symmetric systems [ 43 ] and Floquet dissipative synthetic circuits [ 44 ] , opening new opportunities for engineering the non-equilibrium states of matter in non-Hermitian systems. In typical circuits, inductors exhibit both unavoidable core losses and the Joule heating, leading to energy dissipation in an RLC resonator. A fundamental issue is how such energy loss affects spontaneous \cal{PT} -symmetry breaking. Inspired by this, we investigate, both experimentally and theoretically, the effects of energy dissipation on the coalescence of the eigenmodes in coupled RLC resonators, where energy loss in the inductors is modeled as an effective resistance in series with each inductor. Without considering energy dissipation in the inductors, the system exhibits spontaneous \mathcal{PT} -symmetry breaking as a non-Hermitian parameter crosses the EPs [ 35 , 36 ] . However, when energy dissipation is present, we find that the two real parts of eigenfrequency initially tend to merge as this non-Hermitian parameter increases but eventually develop a gap for sufficiently large values, indicating the breakdown of coalescence. For small values of this non-Hermitian parameter, the voltage of resonators exhibits periodic oscillation with an exponentially decaying amplitude over time. Interestingly, when this parameter exceeds the EPs of a \cal{PT} -symmetric circuit, the amplitude of the oscillation increases exponentially with time. Our above findings provide evidence of the robustness of exponential growth of eigenmodes against unavoidable dissipative perturbations, thereby offering significant implications for non-Hermitian physics. The paper is organized as follows. In Sec. II , we describe the system and show the main results. A summary is presented in Sec. III . II Model and main results The experimental setup consists of two parallel RLC resonators, coupled by a capacitor C_{0} , and integrated on a printed circuit board (PCB) [see Figs. 1 (a) and (c)]. The loss resonator, highlighted in the blue region, has a positive resistance R , while the gain resonator, highlighted in the red region, features a negative resistance -R . The negative resistance is implemented using a voltage-doubling buffer [see Fig. 1 (b)], where the output current follows the relation I_{\text{out}}=(V-2V)/{R}=-{V}/{R} effectively realizing a negative resistance. In our experiment, we use an axial fixed inductor with inductance L=773\,\mu\text{H} , SMD capacitors with capacitances C=4.78\,\text{nF} and C_{0}=1.6\,\text{nF} , and tunable resistors R ranging from 800 \Omega to 15 \text{k}\Omega . The difference in resistance between the positive and negative resistors is less than 0.5\% . The unavoidable magnetic loss in the inductor, accompanied by energy dissipation, disrupts the exact balance between energy gain and loss, causing a deviation from the ideal \cal{PT} -symmetric system. To compensate for this energy dissipation, a gain component is introduced in each coil to realize a \cal{PT} -symmetric circuit [ 35 , 43 , 36 ] . In this work, we introduce a resistance R_{L} in the coil to model the energy dissipation of each inductor. Figure 1: (a) Schematic of \cal{PT} -symmetric RLC resonator. The negative resistance -R in the gain resonator is provided by feedback from a voltage-doubling buffer. The two resonators are coupled by the capacitor C_{0} . (b) Double amplifier with R_{2}=R_{3} . (c) Printed circuit board of the \cal{PT} -symmetric RLC resonator. The buffer consists of high-impedance LF356 voltage followers, designed to isolate the influence of measurement instruments on the RLC resonator. Based on Kirchhoff’s laws, one can obtain the equations for the current I and voltage V of gain and loss nodes \displaystyle I_{n}^{C}+I_{n}^{R}+I_{n}^{L}+I_{n}^{C_{0}}=0\;, (1) \displaystyle V_{n}=R_{L}I_{n}^{L}+L\frac{dI_{n}^{L}}{dt}\;, where the superindices R , L , C , and C_{0} indicate quantities associated with the resistor, inductor, capacitor and coupled capacitor, respectively. The subscripts denote the gain ( n=1 ) and loss ( n=2 ) resonators. Here, the current of the resistors and capacitors take the form I_{n}^{R}=(-1)^{n}V_{n}/{R} , I_{1}^{C_{0}}=C_{0}({dV_{1}}/{dt}-{dV_{2}}/{dt}) , I_{2}^{C_{0}}=-I_{1}^{C_{0}} , and I_{n}^{C}=C{dV_{n}}/{dt} . For brevity, we introduce a set of dimensionless quantities c={C_{0}}/{C} , k={R_{L}}/{R} , \gamma={R}^{-1}\sqrt{{L}/{C}} , and \gamma_{l}=R_{L}\sqrt{{C}/{L}} . Dimensionless time is naturally defined as \tau=\omega_{0}t , where \omega_{0}=1/\sqrt{LC} represents the resonant frequency. Straightforward calculations yield the relations \displaystyle\ddot{V}_{1}= \displaystyle\dot{V}_{1}\frac{\gamma-(1+c)\gamma_{l}}{c+1}+V_{1}\frac{k-1}{c+1}+\ddot{V}_{2}\frac{c}{c+1}+\dot{V}_{2}\frac{c\gamma_{l}}{c+1}\;, (2) \displaystyle\ddot{V}_{2}= \displaystyle-\dot{V}_{2}\frac{\gamma+(1+c)\gamma_{l}}{c+1}-V_{2}\frac{k+1}{c+1}+\ddot{V}_{1}\frac{c}{c+1}+\dot{V}_{1}\frac{c\gamma_{l}}{c+1}\;. where \dot{V}_{i}=dV_{i}/d\tau and \ddot{V}_{i}=d^{2}V_{i}/d\tau^{2} with i=1,2 . By introducing the vector \psi=(V_{1},V_{2},\dot{V}_{1},\dot{V}_{2}) , we can rewrite above equations in analogy to the Schrödinger equation (with \hbar=1 ) i\frac{d\psi}{d\tau}=\text{H}\psi\;, (3) with the Hamiltonian \displaystyle\text{H}=i\begin{pmatrix}0&\mathbb{I}\\ M_{a}&M_{b}\end{pmatrix}\;, (4) where 0 is the zero matrix, \mathbb{I} is the identity matrix, and matrices M_{a} and M_{b} are given by \displaystyle M_{a}=\frac{1}{1+2c}\begin{bmatrix}{(1+c)(k-1)}&-c(k+1)\\ {c(k-1)}&-(1+c)(k+1)\end{bmatrix}\;, (5) and \displaystyle M_{b}=\frac{1}{1+2c}\begin{bmatrix}(1+c)\gamma-(1+2c)\gamma_{l}&-c\gamma\\ c\gamma&-(1+c)\gamma-(1+2c)\gamma_{l}\end{bmatrix}\;. (6) For \gamma_{l}=0 , we can obtain analytically the eigenvalue of the Hamiltonian \omega_{1,3}=\sqrt{\frac{2(1+c)-\gamma^{2}\pm\sqrt{[\gamma^{2}-2(1+c)]^{2}-4(2c+1)}}{2(2c+1)}}\;, (7) and \omega_{2,4}=-\omega_{1,3} . The two positive eigenvalues \omega_{1} and \omega_{3} coalesce at the exceptional point \gamma_{c}=\sqrt{2(1+c)+\sqrt{1+2c}} , beyond which the eigenvalues acquire imaginary components, indicating the breaking of \cal{PT} -symmetry (see Fig. 2 ). Figure 2: Experimentally measured eigenfrequencies \omega=\omega_{r}+i\omega_{i} versus \gamma . The real \omega_{r} and imaginary parts \omega_{i} are shown in (a) and (b), respectively. Solid lines indicate numerical results of \omega for \gamma_{L}=0.0738 . Green-dashed lines are analytical prediction of \omega for \gamma_{L}=0 , manifesting an exponential point at \gamma_{c}=0.29097 . Open circles in (a) are reflections of the experimental data (lower branch) with respect to \omega_{r}\approx 0.864 . Inset in (a): The gap of the real part \Delta\omega_{r} versus \gamma_{L} for \gamma>\gamma_{c} . Red-dashed line indicates the linear fitting, i.e., \Delta\omega_{r}\approx 0.23\gamma_{L} . Open squares in (b) are reflections of the experimental data (upper branch) with respect to \omega_{i}=-0.03706 . In realistic scenarios, nonzero \gamma_{l} arises from the intrinsic energy dissipation of the inductor L . This renders analytical derivation of the eigenvalues for the non-Hermitian Hamiltonian \rm{H} intractable. We therefore perform numerical diagonalization of \rm{H} to resolve its complex eigenvalue \omega=\omega_{r}+i\omega_{i} . Our results show that for very small \gamma_{l} (i.e., \gamma_{l}=0.0738 in Fig. 2 ), the real eigenvalues \omega_{r} have very slight difference from its \gamma_{l}=0 counterpart when \gamma<\gamma_{c} . Intrinsically, however, the two \omega_{r} branches remain non-degenerate at \gamma=\gamma_{c} and develop a finite gap \Delta\omega_{r} for \gamma>\gamma_{c} , signaling the breakdown of spectral coalescence in the \mathcal{PT} -symmetric systems. We find that the gap linearly increases with \gamma_{L} [see the inset in Fig. 2 (a)]. Due to the unavoidable energy dissipation of the coil L , the eigenvalues are complex, i.e., \omega_{j}=\omega_{r,j}+i\omega_{i,j} for j=1,2 [see Fig. 2 (b)]. The imaginary parts of the eigenvalues remain nearly constant and show negligible deviation from each other when \gamma<\gamma_{c} , but they diverge noticeably as \gamma>\gamma_{c} . We further perform experimental measurements of the eigenfrequencies of the coupled RLC resonators. A comparison between \omega and the numerical results is shown in Fig. 2 . One can see that for \gamma<\gamma_{c} , the experimentally measured values of both \omega_{r} and \omega_{i} agree well with numerical simulations based on the model in Eq. ( 4 ). Note that we can extract only one imaginary component of the eigenfrequency because the difference between the two \omega_{i} values is too small. For \gamma>\gamma_{c} , the eigenmode of the electronic circuit associated with a negative imaginary part \omega_{i} decays exponentially. Therefore, only the eigenfrequency with \omega_{i}>0 is experimentally accessible, and it is in perfect agreement with numerical results. Figure 3: (a) Time dependence of the both V_{1} and V_{2} for \gamma=0.0815 (top panels) 0.2545 (middle panels), and 0.3133 (bottom panels). Solid lines indicate our numerical results. We determine the eigenfrequency by applying a Fourier transform to the time evolution of the voltage V_{j} at each node. Figure 3 shows the experimentally measured values of V_{j} for different \gamma . One can see that for \gamma<\gamma_{c} , the V_{j} periodically oscillates with time. The envelope of this oscillation is also periodic, while exhibiting an exponentially decaying amplitude [see Figs. 3 (a) and (b)]. This reveals the existence of two real parts in the eigenfrequency. The rate of the exponential decay of the envelop’s amplitude corresponds to the imaginary parts of the eigenfrequency. For \gamma\approx\gamma_{c} , the amplitude of the envelope of V_{j} is significantly larger compared to smaller values of \gamma . Interestingly, for \gamma>\gamma_{c} , the time evolution of V_{j} exhibits periodic oscillations with a single frequency, and the amplitude increases exponentially with time. In this case, the oscillation frequency and the growth rate of the amplitude are \omega_{r} and \omega_{i} , respectively. For comparison, we also employ the fourth-order Runge-Kutta algorithm to numerically solve Eq. ( 3 ) and obtain the time evolution of V_{j} . Our numerical results can well reproduce the experimentally measured values of V_{j} , confirming the validity of our model. This consistency is clear evidence that our introduction of R_{L} to quantify energy dissipation in the coil L is physically reasonable. Non-Hermiticity is now widely recognized as an important extension of conventional quantum mechanics, providing a framework for describing open quantum systems. Notable phenomena, including the non-Hermitian many-body problem [ 45 ] , biorthogonal dynamical phase transitions [ 46 ] , and non-reciprocal adiabatic tunneling [ 47 ] , have demonstrated the fundamental significance of non-Hermitian physics. To implement \mathcal{PT} -symmetric circuits, gain components should be utilized to compensate for energy loss in inductors [ 35 , 36 ] . In this situation, for \gamma<\gamma_{c} , the voltage of the active RLC resonators exhibits periodic oscillations due to energy gain. However, the unavoidable magnetic loss in inductors, which introduces small energy dissipation, leads to the exponential decay of voltage oscillations, corresponding to a negative value of \omega_{i} . Intrinsically, the eigenmodes of the voltage signal in our system exhibit a transition to exponential growth with time as the non-Hermitian parameter \gamma crosses the EPs, indicating the robustness of exponential energy gain against perturbative dissipation. III Conclusion and discussions The non-Hermiticity of electronic circuits generally arises from dissipative elements that introduce energy loss or gain, breaking energy conservation. In this work, we experimentally realize an electronic circuit where gain and loss RLC resonators are coupled by a capacitor. The gain side is implemented using a voltage-doubling buffer, which provides feedback current with the relation I_{out}=-V/R , effectively achieving a negative resistance. Energy dissipation in the inductors is present in both the gain and loss RLC resonators. Theoretically, we model this energy loss by introducing resistors R_{L} in each inductor, thereby quantifying energy dissipation precisely. For R_{L}=0 , the spontaneous \cal{PT} -symmetry breaking is controlled by a dimensionless parameter \gamma , with the EP occurring at \gamma=\gamma_{c} . We find that V_{j} exhibits periodic oscillations over time, with its amplitude decaying exponentially for \gamma<\gamma_{c} and increasing exponentially for \gamma>\gamma_{c} . Correspondingly, the imaginary part \omega_{i} of the eigenfrequencies is negative for \gamma<\gamma_{c} and positive for \gamma>\gamma_{c} . In addition, the two branches of \omega_{i} diverge as \gamma increases. The two real parts of the eigenfrequency converge as \gamma increases but retain a gap for \gamma>\gamma_{c} . Therefore, the coalescence of eigenmodes is broken by the energy dissipation of inductors. Our finding that \omega_{r} exhibits a gap for \gamma>\gamma_{c} may represent a unique feature of non-Hermitian circuits, whose implications on sensing and wave manipulation warrant further investigation.