Big data, cloud computing and rapidly increasing demand for speed or bandwidth are driving the new generation of Internet which is intelligent, programmable, energy-efficient and secure. The current Internet is not fully transparent and continues to deploy electronic information processing with energy-consuming optical-electrical/electrical-optical conversions1. The long-sought technology for optical information processing (OIP) (or photonic transistor (PT))2 and optical buffering3 are not around the corner yet, and hinders the development of all-optical networks. Moreover, the current Internet is not very secure as it uses light pulses to transmit information across fiber-optic networks. These classical optical pulses can be easily intercepted and copied by a third party without any alert. Quantum Internet4 uses individual quanta of light, i.e., photons to encode and transmit information. Photons can not be measured without being destroyed due to the no-cloning theorem in quantum mechanics5, so any kind of hacking can be monitored and evaded. The realization of quantum Internet requires not only quantum gates and quantum memories for quantum information processing (QIP), but also single-photon transistor (SPT) gated by a single photon to control the flow of information.

The future Internet is very likely the mixture of all-optical Internet with low power consumption and quantum Internet with absolute security. The optical regular Internet would be used by default, but switched over to quantum Internet when sensitive data need to be transmitted. PT and and its counterpart in the quantum limit SPT would be the core components for both OIP and QIP in future Internet. Compared with electronic transistors, PTs/SPTs potentially have higher speed, lower power consumption and compatibility with fibre-optic communication systems.

Several schemes for PT6,7,8,9,10 and SPT11,12,13,14,15,16,17 have been proposed or even proof-of-principle demonstrated. All these prototypes exploit optical nonlinearities, i.e., photon-photon interactions18. However, photons do not interact with each other intrinsically, so indirect photon-photon interactions via electromagnetically induced transparency (EIT)19, photon blockade20 and Rydberg blockade21 were intensively investigated in this context over last two decades in either natural atoms22,23 or artificial atoms including superconducting boxes24,25 and semiconductor quantum dots (QDs)12,13. PT can seldom work in the quantum limit as SPT with the gain greater than 1 because of two big challenges, i.e., the difficulty to achieve the optical nonlinearities at single-photon levels and the distortion of single-photon pulse shape and inevitable noise produced by these nonlinearities26. The QD-cavity QED system is a promising solid-state platform for information and communication technology (ICT) due to their inherent scalability and matured semiconductor technology. But the photon blockade resulting from the anharmonicity of Jaynes-Cummings energy ladder27 is hard to achieve due to the small ratio of the QD-cavity coupling strength to the system dissipation rates12,13,28,29,30,31,32 and the strong QD saturation33. Moreover, the gain of this type of SPT based on the photon blockade is quite limited and only 2.2 is expected for In(Ga)As QDs12,13.

In this work, a different PT and SPT scheme exploiting photon-spin interactions rather than photon-photon interactions is proposed based on a linear quantum-optical effect - giant optical Faraday rotation (GFR) induced by a single QD-confined spin in a single-sided optical microcavity34. This spin-cavity transistor is genuinely a quantum transistor in three aspects: (1) it is based on a quantum effect, i.e., the linear GFR; (2) it has the duality as a quantum gate for QIP and a classical transistor for OIP; (3) it can work in the quantum limit as a SPT to amplify a single-photon state to Schrödinger cat state. Therefore this new-concept transistor can be more powerful than the traditional electronic transistors. Theoretically the maximum gain can reach ~105 in the state-of-the-art pillar microcavity, several orders of magnitude greater than previous PT/SPT schemes6,7,8,9,10,11,12,13,14,15,16,17. The large gain is attributed to the linear GFR that is robust against classical and quantum fluctuations and the long spin coherence time compared with the cavity lifetime. The maximal speed which is determined by the cavity lifetime has the potential to break the terahertz (THz) barrier for electronic transistors35,36. Based on this versatile spin-cavity transistor, optical Internet1, quantum computers (QCs)37,38 (either spin-cavity hybrid QCs or all-optical QCs), and quantum Internet4 could become reality even with current semiconductor technology.

Results and Discussions

Linear GFR for robust and scalable quantum gates

A single electron (or hole) spin confined in a charged QD in an optical microcavity can induce GFR34 – a kind of optical gyrotropy (or optical activity)39. Although cavity QED systems are typically highly nonlinear, it was recently found that GFR exhibits both nonlinear and linear behaviors33. The nonlinear GFR is sensitive to the power of incoming light occurs when the QD is partially saturate, and has been demonstrated experimentally for single QD spin40,41,42,43,44, QD spin ensemble45, and single atom46,47,48 in the weak coupling regime of cavity QED or without a cavity. The linear GFR is independent of the power of incoming light and occurs when the QD is pinned to the ground state within the non-saturation window (NSW) or in the weak-excitation limit. The linear phase shift was reported recently in strongly-coupled atom-cavity systems49,50. The linear GFR in QD-cavity systems is responsible for both robust quantum gate operations (in this subsection) and transistor operations (in the next subsection) although it has not been demonstrated yet.

Figure 1(a) shows such a spin-cavity unit with a negatively charged QD in a single-sided pillar microcavity. This type of cavity can be fabricated from a planar microcavity defined by two distributed Bragg reflectors (DBRs) with the cavity length chosen to be one wavelength λ such that the field maxima is situated in the middle of the planar microcavity. The front mirror is made partially reflective and the back mirror 100% reflective, allowing unit cavity reflection if the cavity side leakage is negligibly small. Three-dimensional confinement of light is provided by the two DBRs and the transverse index guiding. The cross section of the micropillar is made circular in order to support circularly polarized light besides linearly polarized light. Some photonic crystal nanocavities with specific spatial symmetry51 could support circularly polarized light and are suitable for this work, too. However, the advantage to use the pillar microcavity is the high coupling efficiency as the fundamental cavity mode which is gaussian-like can match perfectly with the external laser beam.

Figure 1: Structure of the spin-cavity unit.
figure 1

(a) A charged quantum dot is embedded in a single-sided pillar microcavity with one end mirror partially reflective and another end mirror 100% reflective allowing unit cavity reflectance when the cavity side leakage is neglected. The circular pillar cross-section supports circularly polarized light. (b) Optical transitions in a negatively-charged quantum dot follow spin selection rules: a photon in the |L〉 state couples to the transition |↑〉 ↔ |↑↓〉 only, whereas a photon in the |R〉 state couples to the transition |↓〉 ↔ |↓↑〉 only due to the conservation of angular momentum and the Pauli exclusion principle.

A negatively (or positively) charged QD has an excess electron (or hole) confined in the QD. Charging a QD can be achieved via modulation doping, tunneling52, or optical injection. The ground states of the charged QD are the electron (or hole) spin states, and the excited states are the spin states of the negatively charged exciton X (or positively charged exciton X+) [see Fig. 1(b)]. In the absence of external magnetic field, both the ground and excited states of charged QD are two-fold degenerate due to the Kramers theorem. The electron spin degeneracy could be lifted by the nuclear spin magnetic fields53 via the electron-nucleus hyperfine interactions in In(Ga)As QDs, however, the Zeeman splitting is too small to spoil the linear GFR33. The hole spin degeneracy is not affected by the nuclear spin fields due to the lack of the hole-nucleus hyperfine interactions.

Due to the conservation of total spin angular momentum and the Pauli exclusion principle, the left circularly polarized photon (marked by |L〉 or |σ+〉) only couples to the transition |↑〉 ↔ |↑↓〉, and the right circularly polarized photon (marked by |R〉 or |σ〉) only couples to the transition |↓〉 ↔ |↓↑〉 [see Fig. 1(b)]. Here |↑〉 and |↓〉 represent electron spin states , |〉 and |〉 represent heavy-hole spin states with the spin quantization axis z along QD growth direction, i.e., the input/output direction of light. The weak cross transitions due to the heavy-hole-light-hole mixing54 can be corrected and are neglected in this work. Note that the photon polarizations are marked by the input states to avoid any confusion due to the temporary polarization changes upon reflection.

If the spin is in the |↑〉 state, a photon in the |L〉 state can couple to the QD and feels a “hot” cavity, whereas a photon in the |R〉 state can not couple to the QD and feels a “cold” cavity [see Fig. 1(b)]. If the spin is in the |↓〉 state, a |R〉-photon feels a “hot” cavity and a |L〉-photon feels a “cold” cavity.

Although the incoming light is fully reflected back from a single-sided cavity, there exists significant phase difference between the reflection coefficients of the “hot” and “cold” cavity as the QD-cavity interactions can strongly modify the cavity properties. This cavity-QED effect is verified by the calculations using two approaches (see Methods): an analytical method by solving Heisenberg-Langevin equations of motions in the semi-classical approximation, and a numerical but exact method by solving master equation with a quantum optics toolbox55,56. The calculated results using these approaches can be found in our previous work33.

The phase difference between the “hot” and “cold” cavity can be mapped to that between left-handed and right-handed circularly polarized light. This results in the optical rotation of polarization of a linearly polarized light after reflection, i.e., GFR effect which is induced by a single spin. GFR can be regarded as a macroscopic imprint of the microscopic spin selection rules for optical transitions in charged QD (sometimes it is also called Pauli blockade) [see Fig. 1(b)]. GFR is a type of magnetic optical gyrotropy (also known as magnetic optical activity) in the presence of magnetic field or magnetization39. A key feature of GFR is its spin tunability, which makes the spin-cavity unit versatile for QIP and OIP. Another merit is the linearity of GFR that remains constant with increasing the power of incoming light.

The linear GFR occurs around the cavity resonance in the strong coupling regime g  (κ + κs,γ) or in the Purcell regime γ < 4g2/(κ + κs) < (κ + κs) when the input power is less than Pmax (see Methods for its definition) such that the QD stays in the ground state. Within the NSW or in the weak-excitation limit, the coherent scattering dominates the reflection process and the semi-classical approximation yields the same results as the full quantum model33. Taking 〈σz〉 = −1, the steady-state reflection coefficient can be obtained from Eq. (12) in Methods,

where ω, ωc, are the frequencies of incoming light, cavity mode, and the X transition, respectively. g is the QD-cavity coupling strength which is given by (if the QD is placed at the maxima of cavity field) with f being the X oscillator strength (f:10–100 for InAs- or GaAs-based QDs) and Veff being the cavity mode volume. κ/2 is the the cavity field decay rate into the input/output port, and κs/2 is the side leakage rate of the cavity field including the material background absorption. γ/2 is the total QD dipole decay rate including the spontaneous emission rate into leaky modes and the pure dephasing rate γ*, i.e., . The pure dephasing rate can be neglected when the QD stays in the ground state as being proved in recent experiments on generation of high-quality single photons from In(Ga)As QDs under weak resonant excitation57,58. For pillar microcavity, the spontaneous emission rate into leaky modes is approximately equal to the free-space emission rate as the reduced density of leaky modes can be compensated by the Purcell enhancement due to the light confinement from the planar cavity.

For convenience to discussions, the resonant condition is assumed in this work although this assumption is not necessary as the linear GFR is tolerable to the frequency mismatch . In the one-dimensional atom regime where 4g2 (κ + κs)γ (this includes the strong coupling regime and part of the Purcell regime), Eq. (1) yields and when the light frequency ω lies within the NSW, i.e., |ω − ω0| < −gσz〉 where 〈σz〉 depends on the intensity of the incoming light. If the cavity side leakage is smaller than the input/output coupling rate, i.e., κsκ, and ϕ0(ω)  [−π, +π] can be achieved for the cold cavity. The spin-dependent (conditional) phase shift can be used to build a deterministic photon-spin entangling gate (precisely speaking, a conditional phase gate or a parity-check gate) with the phase shift operator defined as ref. 34

where the phase shift Δϕ = ϕh − ϕ0 is chosen to ±π/2 by setting the frequency detuning Δω = ω − ω0 ≈ ±(κ + κs)/2. Higher gate fidelity can be achieved in strongly coupled QD cavity QED systems which has been demonstrated in various micro- or nano-cavities12,13,28,29,30,31,32. In the state-of-the-art pillar microcavities28,59, g/(κ + κs) = 2.4 is achieved for In(Ga)As QDs and this parameters is used for judging the performance of transistor in this work.

As the linear GFR occurs when the QD stays in the ground state, this photon-spin entangling gate is robust against quantum fluctuations33 either from the inside of cavity or from the outside, e.g., the intensity fluctuations of the incoming light. The linear GFR within the NSW is also resistant to spectral diffusion, pure dephasing, charge and nuclear spin noise, high-order dressed state resonances (i.e., the higher manifolds of the Jaynes-Cummings ladder), and small QD-cavity mismatch which could be induced by external electric/magnetic fields, all of which could occur in realistic QDs. This universal, deterministic and robust photon-spin entangling gate is promising for scalable solid-state QIP.

It is worth noting that rather than the π/2 phase shift used here, the conditional π phase shift can be also used for a controlled phase flip gate in a strongly coupled atom-cavity system60. Since the pioneer work by Kimble’s group46, both nonlinear47,48 and linear49,50 phase shift of π (and π/2 by frequency detuning) have been demonstrated in atomic cavity-QED systems recently. Significant progress has also been made in QD-cavity systems, e.g., the nonlinear GFR of ~6° induced by a single hole spin43 or electron spin44, a photon sorter61, and a quantum phase switch62. However, all these experiments were performed in the weak coupling regime of cavity QED63,64 (equivalent to waveguide-QED65) where the nonlinear GFR (or the nonlinear phase shift) dominates and the GFR bandwidth is quite small (limited by the QD linewidth). The linear GFR in strongly-coupled QD-cavity systems is desired as it occurs with a broader bandwidth (limited by the cavity mode linewidth) and allows both robust quantum gate and transistor operations.

For the sake of convenience, some properties of the photon-spin phase shift operator are listed below


where and are linear polarization states, and are diagonal linear polarization states. It is worth noting that |R〉|↑〉, |L〉|↑〉, |R〉|↓〉, and |L〉|↓〉 are eigen states of the spin-cavity hybrid device. The linearly polarized light will turn its polarization by ±45° after reflection (i.e., GFR effect) if the spin is set to |↑〉 or |↓〉. These properties are used in this work.

By itself the spin-cavity unit can be used to initialize the spin via single-photon based spin projective measurement together with classical optical pulses injected from the cavity side. Assume the spin is in a unknown state α|↑〉 + β|↓〉, and the incoming photon in the |H〉 state. After the photon-spin interaction, the photon and spin become entangled, i.e., α|A〉|↑〉 + β|D〉|↓〉, On detecting the photon in |A〉, the spin is projected to |↑〉. On detecting the photon in |D〉, the spin is projected to |↓〉. To convert the spin from |↓〉 to |↑〉 or vice versa, a spin rotation of π around y axis is required (see Fig. 2(a) for the definition of x, y, z axes). This can be achieved using a ps or fs optical (π)y pulse (injected from the cavity side66) via the optical Stark effect67. To prepare the superposition state such as , an optical pulse can be applied. With these techniques, the electron spin in a pillar microcavity can be prepared to an arbitrary state deterministically.

Figure 2: Diagram of the spin-based photonic transistor.
figure 2

(a) Firstly, the spin is initialized to , and the gate photon is prepared in an arbitrary state |ψph〉 = α|H〉 + β|V〉; Secondly, the gate photon state is transferred to spin by measuring the photon in the {|D〉, |A〉} basis after reflection; Thirdly, the spin state is transferred to N photons injected from the source S by measuring the spin in the {|+〉, |−〉} basis. As a result, an arbitrary quantum state of a gate photon is transferred (or “amplified”) to the same state encoded on N photons in the channel. PBS (polarization beam splitter), D1-D2 (single photon detectors). Note that the PBS is oriented in the {|D〉, |A〉} basis, and transmits the photon in state |A〉 and reflects the photon in state |D〉. (b) The maximum gain – speed product as a function of the QD-cavity coupling strength g. The spin coherence time is taken as T2 = 1 μs (see the text).

Spin coherence time T2 is an important parameter for both quantum gate and transistor operations. In GaAs-based or InAs-based QDs, the electron spin dephasing time can be quite short (~ns) due to the hyperfine interaction between the electron spin and 104 to 105 host nuclear spins53. To suppress the nuclear spin fluctuations, spin echo or dynamical decoupling techniques68 could be applied to recover the electron spin coherence using optical pulses67 and/or single-photon pulses69. Based on the spin echo techniques, the electron spin coherence time T2 = 1μs has been reported recently in a single In(Ga)As QD67, which is taken in this work to estimate the transistor gain-speed product (see Fig. 2(b)). Note that the spin echo technique is compatible with the quantum gate and transistor operations in the spin-cavity unit.

Aside from the deterministic photon-spin entangling gate, this spin-cavity unit can also work as deterministic photon-photon, spin-spin entangling gates and deterministic photon-spin interface or heralded spin memory, single-shot quantum non-demolition (QND) measurement of single spin or photon, complete Bell-state generation, measurement and analysis as well as on-chip quantum repeaters34,69,70,71,72. Assisted by single photon or single spin gate operations, these gates can be converted to the popular quantum gates in different systems, e.g., controlled-NOT (CNOT) or hyper CNOT gates, Toffoli gates, and Fredkin gates which can be useful for scalable quantum computing73,74,75,76. Similar quantum gates, i.e., the controlled phase-flip gates were proposed by Duan and Kimble60 in atom-cavity systems have been experimentally demonstrated recently49,50. However, the atomic systems require complicated and expensive equipment for cooling and trapping and the scalability is really a big challenge from a practical point of view. Moreover, in atomic systems these quantum gates work in the MHz range due to the small atom-cavity coupling strength. The QD-cavity systems have inherent scalability, high speed (tens to hundreds of GHz) due to the larger QD-cavity coupling strength, and the ease to fabricate with matured semiconductor technology, thus providing an ideal platform for solid-state QIP.

Linear GFR for photonic transistor

The spin-dependent linear GFR can be utilized to make a spin-based SPT as shown in Fig. 2(a). A gate photon sets the spin state via projective measurement and controls the polarization of light in the photonic channel between the source S and the drain D.

The spin is initialized to using single-photon based spin projective measurement in combination with an ultrafast optical pulse injected from the cavity side. The gate photon is prepared in an arbitrary state |ψph〉 = α|H〉 + β|V〉. After the photon interacting with the spin, the joint state becomes

The photon is then measured in the {|D〉, |A〉} basis. On detecting the photon in |A〉 (a click on D1), the spin is projected to |ψs〉 = α|↑〉 + β|↓〉. On detecting the photon in |D〉 (a click on D2), the spin is projected to |ψs〉 = α|↓〉 − β|↑〉, which can be converted to |ψs〉 = α|↑〉 + β|↓〉 by spin rotations. The spin-cavity unit works as a photon-spin quantum interface, with which the photon state is transferred to spin state deterministically.

Assume there are N photons in the |H〉 state from the source S (the photon number N is the maximum gain which will be discussed later). After all photons interacting with the spin in sequence, the joint state becomes

An optical pulse is then applied from the cavity side to perform the spin Hadamard transformation. After that, a photon in the |H〉 state is sent to perform the spin measurement, and the joint state are projected to a superposition state

where “+” is taken for spin |+〉 and “−” for spin |−〉. The negative sign can be converted to the positive using a phase shifter.

As a result, an arbitrary quantum state of a single gate photon is transferred or “amplified” to the same state encoded on N photons, which is of Greenberger-Horne-Zeilinger (GHZ) state like or Schrödinger-cat state like77. In this sense, this SPT is genuinely a quantum transistor based on a quantum effect – GFR, and exhibits the dual nature as a quantum gate and a transistor. As the original state of the gate photon is destroyed after the transistor operation, this SPT does not violate the no-cloning theorem in quantum mechanics5. This transistor could generate entanglement of hundreds to thousands of photons and has the potential to break the current record of 10-photon entanglement78. The multiphoton entanglement are essential to quantum communications79 and quantum metrology80. Previous calculations of the entanglement fidelity and efficiency in terms of single-photon transportation34,70 can be extended to the case of n-photon Fock state (n ≤ Pmax) as long as the linear GFR preserves when the incoming light power is less than Pmax. This is because the n-photon transitions for the excitation of the nth manifold of dressed states do not affect the linear GFR occurring within the NSW, and the reflection amplitude and phase remain the same for both single-photon and n-photon Fock states33. The efficiency and fidelity to generate the N-photon GHZ- or cat-like states in Eq. (7) depends on cavity QED parameters (g, κ, κs, γ), the spin coherence time T2, the time interval between photons as well as the precision for spin manipulations69.

The time interval between the channel and gate photons should be less than by the spin coherence time T2 which defines the time window for the transistor operation. The photon rate in the channel can go as high as Pmax where the linear GFR preserves. As the spin state can be controlled by a single photon, the maximum gain of the SPT is exactly the maximum photon number allowed in the channel, i.e.,

where τ is the cavity lifetime which determines the cutoff frequency fT of SPT (i.e., the maximum speed). The maximum gain-speed product is thus

where ħ is the reduced Planck constant. The gain-speed product increases with increasing the coupling strength g or the spin coherence time T2 [Fig. 2(b)]. In the state-of-the-art pillar microcavity28,59 where g/(κ + κs) = 2.4 (g = 80 μeV, κ + κs = 33 μeV), the maximum gain can reach 7 × 104, which surpasses all other SPT protocols by several orders of magnitude11,12,13,14,15,17. However, the high gain is at the cost of a low speed (fT ~ 50 GHz in this case) and vice versa. In order to raise the speed, the cavity lifetime can be reduced. For example, if the cavity decay rate increases to 660 μeV, fT goes up to 1 THz with the gain down to 3.5 × 103. However, too much cavity decaying will wash out GFR34 if 4g2/(κ + κs) < γ, and causes the failure of the SPT operation.

Besides its quantum nature, the SPT can also work as a classical PT: a gate photon in a mixed state of |H〉 and |V〉 can be amplified to the same mixed state of N photons. Moreover, this classical PT also works if a gate photon is replaced by a classical optical pulse as long as the optical power is much less than Pmax. It is worth noting that this spin-cavity transistor satisfies most criteria for all-optical transistors that could compete with the electronic counterparts2, e.g., logic level independent of loss due to the linear GFR which is immune to the power variations of incoming light.

Different from the spin-based electronic transistors exploiting the Rashba spin-orbit interactions81, this spin-based PT exploiting the spin-cavity interactions does not suffer from the limitation from the RC time constants and the transit time, so it has the potential to break the THz barrier for all electronic transistors including the state-of-the-art high electron mobility transistors (HEMTs)35,36. Moreover, the PT consumes less energy as there exists neither Joule heating nor capacity charging in the photonic channel.

Similar to the conventional transistor, this spin-cavity transistor can be used as photonic switches or modulators with high-speed (up to THz). The inclusion of a spin into the cavity breaks the time inversion symmetry if the spin orientation is fixed. This results in the optical non-reciprocity39 which can be utilized for making various non-reciprocal photonic devices such as diodes, isolators and circulators. All these devices are useful for quantum networking and all-optical networking.

Linear GFR for photonic router

In analogy to classical routers in regular Internet, which direct the data signal to its intended destination according to control information contained in the IP address, quantum routers82 are a key building block in quantum networks and quantum Internet, which direct a signal quantum bit (qubit) to its desired output port determined by the state of a control qubit, but keeping the state of the signal qubit unaltered.

The spin-cavity unit is an ideal component to make a quantum router (Fig. 3). The photon is used as the signal qubit encoded to an arbitrary state |ψs〉 = α|H〉 + β|V〉 to be directed to its destination (port c or d), and the spin is used as the control qubit.

Figure 3: Diagram of the spin-based photonic router.
figure 3

The control spin directs the signal photon to: (a) port c; (b) port d; (c) superposition of two modes in port c and d. PBS1, PBS2 (polarization beam splitters), λ/2 (wave plate). Note that PBS1 is set to the {|H〉, |V〉} basis and PBS2 to the {|D〉, |A〉} basis.

If the control spin is set to |ψc〉 = |↓〉 [Fig. 3(a)], the signal photon state is changed from |ψs〉 = α|H〉 + β|V〉 to |ψs〉 = α|D〉 + β|A〉 after the photon-spin interaction. The photon state is then split by PBS1 and combined by PBS2, and finally the signal photon comes out from port c with its state unchanged, i.e, |ψs〉 = α|Dc + β|Ac.

If the control spin is set to |ψc〉 = |↑〉 [Fig. 3(b)], the signal photon state is changed from |ψs〉 = α|H〉 + β|V〉 to |ψs〉 = α|A〉 − β|D〉 after the photon-spin interaction. After PBS1, PBS2, and a λ/2 wave plate, the signal photon comes out from port d with the state |ψs〉 = α|Dd + β|Ad, the same as the original one.

If the control spin is set to |ψc〉 = γ|↑〉 + δ|↓〉, after the photon-spin interaction the joint output state becomes

which is generally the superposition of two modes in port c and d [Fig. 3(c)]. Fully controlled by the spin state, the signal photon can be directed to port c, or port d, or the superposition of port c and d, whereas the signal photon state remains unchanged. As the spin-cavity unit can work as a deterministic photon-spin interface as discussed above, the control spin can be replaced by a photon, such that the quantum router becomes fully transparent. It is worthy noting that the PBS1 and PBS2 just split and recombine the polarization states, rather than forming an optical interferometer. Thus the phase instability is not an issue for this router as interference does not play a role here. Compared with another quantum router scheme based on linear optics82, this spin-cavity quantum router is deterministic and scalable to multiple-photon routing.

Aside from single-photon quantum router, the spin-cavity unit can also work as a classical photonic router if single photons are replaced by classical optical pulses as long as the light power is below Pmax. As the Joule heating and capacity charging are both absent in the photonic channel, photonic routers consume less energy and would replace the electronic routers in future energy-efficient Internet.

It is worth pointing out that the spin-cavity unit can also be used to route light (or photons) carrying orbital angular momentum (OAM)83 if OAM is converted to circular polarization of light or photons via a q-plate84.

Conclusions and outlook

The spin-based high-gain and high-speed PT/SPT and related devices discussed above can be made in parallel based on giant circular birefringence (GCB) in another type of spin-cavity unit with a double-sided microcavity85,86. In both cases, the maximum transistor gain could reach ~105 in the state-of-the-art pillar microcavity depending on the QD-cavity coupling strength and the spin coherence time. The maximal speed which is determined by the cavity lifetime could reach tens to hundreds of GHz and has the potential to break the THz barrier for electronic transistors.

An unusual feature of these spin-cavity units is the duality as quantum gates and transistors thanks to the linear GFR/GCB which are robust against quantum and classical fluctuations including the intensity variations of input light, spectral diffusion and pure dephasing in QDs, charge and spin noise in QDs, and even electric/magnetic fields. On the one hand, the spin-cavity units work as universal, deterministic and robust quantum gates for QIP. On the other hand, the spin-cavity units work as optical transistors for OIP. This work demonstrates that the spin-cavity units provide a solid-state platform ideal for future green and secure Internet - a combination of all-optical Internet1 with quantum Internet4, which is very likely to happen within the next 10–20 year timescale. This work series opens a new research field - spin photonics which studies the physics and applications of a single QD spin in different photonic structures including cavities, waveguides, chiral structures, and topological structures.


Semiclassical model

The Heisenberg equations of motions87 for the cavity field operator and the QD dipole operators σ, σz, together with the input-output relation88 can be written as

where all the parameters here have the same definitions and meanings as in Eq. (1).

If the correlations between the cavity field and the QD dipole are neglected (this is called the semiclassical approximation), and . The semiclassical approximation can be applied in three cases: (1) low-power limit where the QD stays the ground state (weak-excitation approximation); (2) high-power limit where the QD is saturated; (3) within the NSW where the QD is pinned to the ground state33. The reflection coefficient can thus be derived as

The population difference 〈σz〉 is given by

and the average cavity photon number by

where is the critical photon number which measures the average cavity photon number required to saturate the QD response, and nc = 2.2 × 10−4 is taken in this work. is the input light power. 〈σz〉 is the QD population difference between the excited state and the ground state, and can be used to measure the saturation degree. 〈σz〉 ranges from −1 to 0. If 〈σz〉 = −1, the QD is in the ground state (not saturated); if 〈σz〉 = 0, QD is fully saturated, i.e., 50% probability in the ground state and 50% probability in the excited state. If 〈σz〉 takes other values, the QD is partially saturated.

By solving Eqs (13) and (14), 〈σz〉 and 〈n〉 can be obtained at any input power. Note that 〈σz〉 and 〈n〉 are dependent on the input power, frequency ω and coupling strength g. Putting 〈σz〉 into Eq. (12), the reflection coefficient can be derived.

The linear GFR persists as long as the NSW between the first-order dressed state (or polariton state) resonances is open33. From Eqs (13) and (14), it can be estimated that the NSW is closed roughly at when 〈σz〉 = −1/2. The higher the coupling strength g, the higher powers the linear GFR can preserve.

Full quantum model - master equation

The reflection coefficient can be also calculated numerically in the frame of master equations in the Lindblad form87 using a quantum optics toolbox in MATLAB55 or PYTHON56. The master equation for the spin-cavity system can be written as

where the parameters are defined in the same way as in Eq. (11), is the Liouvillian and HJC is the driven Jaynes - Cummings Hamiltonian with the input field driving the cavity. In the rotating frame at the frequency of the input field, HJC can be written as

where the input field is associated with the output field and the cavity field by the input-output relation88, .

Although an analytical solution to the master equation in Eq. (15) is very difficult, the quantum optics toolbox55,56 provides a numerical calculation of the density matrix ρ(t). By taking the operator average in the input-output relation, the steady-state reflection coefficient can be calculated using the following expression

Additional Information

How to cite this article: Hu, C. Y. Photonic transistor and router using a single quantum-dot-confined spin in a single-sided optical microcavity. Sci. Rep. 7, 45582; doi: 10.1038/srep45582 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.