08 Jun 2026
Two Coupled Quantum Oscillators V: Circuit Depth as Quantum Complexity
This is Part V of the coupled-oscillator series. In Part IV, we saw that synchronization and mutual information can separate: one measures coordinated motion, while the other measures total shared information. The next question is operational:
How costly is it to prepare such a correlated quantum state?
The paper answers this using circuit depth as a measure of quantum complexity.
1. What Circuit Depth Means
A quantum circuit is a sequence of gates. If a state $|\psi_R\rangle$ is transformed into another state $|\psi_T\rangle$, then one writes
\[\|\psi_T\rangle=U\|\psi_R\rangle.\]The unitary $U$ can be decomposed into elementary gates:
\[U=O_nO_{n-1}\cdots O_2O_1.\]The circuit depth counts how many sequential layers of gates are needed. A deeper circuit usually means:
| Larger depth means | Consequence |
|---|---|
| more operations | more resources |
| longer implementation | more exposure to noise |
| more entangling structure | richer target state |
| harder preparation | larger complexity |
In the Nielsen geometric approach, complexity is treated as a distance: how far the target state is from a chosen reference state, when one is allowed to move using a specified gate set.
2. Reference State and Target State
The reference state is chosen to be a simple factorized Gaussian:
\[\psi_R(x_1,x_2) = \sqrt{\frac{m\omega_R}{\pi}} \exp\left[ -\frac{m\omega_R}{2}(x_1^2+x_2^2) \right].\]This is the ground state of two identical, decoupled oscillators with reference frequency $\omega_R$.
It is called a reference state because it is easy to prepare. It has no mixed term $x_1x_2$, so it contains no direct correlation between the two coordinates.
The target state is the time-evolved Gaussian derived in Part II:
\[\psi_T(x_1,x_2) = \mathcal N \exp\left[ -\frac{1}{2} \left(A_1x_1^2+A_2x_2^2-A_{12}x_1x_2\right) \right].\]The target state is harder because it can have:
| Coefficient | Meaning |
|---|---|
| $A_1$ | width/frequency scale of the first coordinate |
| $A_2$ | width/frequency scale of the second coordinate |
| $A_{12}$ | cross-correlation or entangling structure |
The circuit has to convert the simple reference Gaussian into this more structured target Gaussian.
3. What Gates Are Needed?
For Gaussian states, one does not need arbitrary gates. Gaussian transformations can be generated by quadratic combinations of positions and momenta.
The paper uses two types of gates.
Local Scaling Gates
The local scaling gates are
\[O_{jj} = \exp\left[ \frac{i\epsilon}{2}(x_jp_j+p_jx_j) \right].\]These gates squeeze or scale one oscillator. They change the width of the Gaussian in the $x_j$ direction.
In simple language:
\[O_{jj}:\quad \text{change width of mode }j.\]Entangling Gates
The nonlocal gates are
\[O_{ij} = \exp(i\epsilon x_ip_j), \qquad i\neq j.\]These gates mix the two oscillators. They are needed when the target Gaussian contains the cross term
\[A_{12}x_1x_2.\]In simple language:
\[O_{ij}:\quad \text{create correlation between the two modes}.\]4. Why Gaussian Complexity Is Tractable
Both the reference state and target state are Gaussian. A Gaussian state is fully determined by a quadratic form in the exponent. Therefore the preparation problem reduces to a matrix problem: how much scaling and mixing is required to transform one quadratic form into another?
The reference quadratic form is proportional to
\[\omega_R \begin{pmatrix} 1 & 0\\ 0 & 1 \end{pmatrix}.\]The target quadratic form can be represented by a symmetric matrix. With the exponent convention used in the previous post,
\[G_T= \begin{pmatrix} A_1 & -A_{12}/2\\ -A_{12}/2 & A_2 \end{pmatrix},\]if the exponent is written as
\[-\frac{1}{2}(A_1x_1^2+A_2x_2^2-A_{12}x_1x_2).\]The diagonal entries measure local scaling. The off-diagonal entry measures mixing.
Convention note: Some papers absorb the factor of $2$ into the definition of the cross coefficient. The circuit-depth formula below is written in the paperβs $A_{12}$ convention. If one uses the strict matrix entry $-A_{12}/2$, corresponding factors of $2$ must be moved consistently.
5. Closed Formula for Circuit Depth
The paper gives the circuit depth as
\[\mathcal D(U) = \frac{1}{2} \log\left[ \frac{A_1A_2-A_{12}^2}{\omega_R^2} \right] + \left|\frac{A_{12}}{A_1}\right|.\]This formula has two parts:
| Term | Meaning |
|---|---|
| $\frac{1}{2}\log\left[\frac{A_1A_2-A_{12}^2}{\omega_R^2}\right]$ | global scaling cost |
| $\left|\frac{A_{12}}{A_1}\right|$ | entangling or mode-mixing cost |
The logarithmic part compares the overall size of the target Gaussian to the reference Gaussian. The second part measures how much cross-correlation must be created.
So the complexity grows for two reasons:
- the target widths differ from the reference width,
- the target state has nonzero coupling between $x_1$ and $x_2$.
6. Meaning of the Determinant Term
The combination appearing in the paper,
\[A_1A_2-A_{12}^2\]plays the role of an effective determinant-like scale of the target Gaussian structure in that convention. It measures the total area or volume scale of the Gaussian in the two-dimensional coordinate space.
If the target Gaussian is very different from the reference Gaussian, this determinant changes significantly, and the logarithmic term grows.
If $A_{12}=0$, the depth becomes
\[\mathcal D(U) = \frac{1}{2}\log\left(\frac{A_1A_2}{\omega_R^2}\right).\]This is pure scaling complexity. No entangling gate is needed.
7. Meaning of the Entangling Term
The term
\[\left|\frac{A_{12}}{A_1}\right|\]is the additional cost caused by the mixed Gaussian term. If $A_{12}$ is large, the target state has strong mode mixing. Preparing it from an uncorrelated reference state requires nonlocal gates.
This is the operational meaning of the cross coefficient:
\[A_{12}\neq0 \quad\Longrightarrow\quad \text{entangling cost}.\]Thus the same coefficient that signals correlation in the wavefunction also increases circuit depth.
8. Weak Coupling Regime
In the weak coupling regime,
\[g\ll \omega_1,\omega_2,\omega_c,\]the mixing angle is small and the cross coefficient $A_{12}$ is suppressed. The target state is close to a product Gaussian.
Then
\[\left|\frac{A_{12}}{A_1}\right|\approx0,\]and the depth reduces to
\[\mathcal D(U) \approx \frac{1}{2} \log\left[ \frac{A_1A_2}{\omega_R^2} \right].\]This regime is close to adiabatic. The target state mainly differs from the reference state by local scaling, not by strong entanglement.
| Feature | Weak coupling behavior |
|---|---|
| $A_{12}$ | small |
| entangling cost | small |
| circuit depth | mostly logarithmic |
| physical behavior | low-complexity preparation |
9. Strong Coupling Regime
In the strong coupling regime, the modes mix significantly. Near maximal mixing,
\[\theta\approx\frac{\pi}{4}.\]The paper states that the interference term is amplified, approximately as
\[A_{12}\approx\sqrt{\frac{g}{2}}.\]Then the circuit depth becomes
\[\mathcal D(U) \approx \frac{1}{2} \log\left[ \frac{A_1A_2-g/2}{\omega_R^2} \right] + \frac{\sqrt{g/2}}{A_1}.\]The additive term now grows with coupling strength. This indicates that the system is no longer just being locally scaled. It is being strongly mixed.
| Feature | Strong coupling behavior |
|---|---|
| mixing angle | approaches $\pi/4$ |
| $A_{12}$ | grows |
| entangling gate cost | important |
| circuit depth | increases |
| physical behavior | nonadiabatic/high-complexity regime |
10. Magnetic-Field-Dominated Regime
When the magnetic field dominates,
\[\omega_c\gg \omega_1,\omega_2,g,\]the diagonal frequency shifts become large. The effective local frequencies dominate over the off-diagonal coupling.
In this regime, the paper obtains the simple scaling
\[\mathcal D(U) \approx \log\left(\frac{\omega_c}{\omega_R}\right).\]This says that the main cost is no longer entangling the two modes. The main cost is scaling the reference state to match the large magnetic-field-induced frequency scale.
Interpretation: A strong magnetic field can suppress mode entanglement in the complexity formula, but it still changes the local frequency scale. The circuit depth therefore survives as a logarithmic scaling cost.
11. Adiabatic and Nonadiabatic Regimes
The paper separates the dynamics into two broad regimes.
| Regime | Description | Complexity behavior |
|---|---|---|
| adiabatic | smooth, ground-state-following evolution | low depth, mostly logarithmic |
| nonadiabatic | rapid transitions and strong mode mixing | high depth, entangling term important |
The crossover is controlled by:
\[g,\qquad \omega_c,\qquad \Delta=\omega_1^2-\omega_2^2.\]The detuning $\Delta$ is especially important. If the two oscillator frequencies are nearly equal, then
\[\Delta\to0.\]Near this resonant condition, small parameter changes can strongly affect the mixing angle. This makes control more delicate and can increase circuit depth sharply.
12. Experimental Meaning
The circuit-depth formula gives practical guidance.
If one wants high-fidelity, low-complexity control, then one should avoid unnecessary strong mode mixing. This suggests:
| Desired goal | Helpful regime |
|---|---|
| low complexity | weak coupling |
| stable control | avoid near-degenerate detuning |
| lower entangling overhead | suppress large $A_{12}$ |
| simple scaling behavior | field-dominated logarithmic regime |
If one wants large correlations, then strong coupling may help. But the price is higher circuit depth and greater sensitivity to errors.
This is the trade-off:
\[\text{more correlation} \quad\Longrightarrow\quad \text{larger preparation cost}.\]13. Connection With Previous Parts
The formula for circuit depth depends directly on the Gaussian coefficients from Part II:
\[A_1,\qquad A_2,\qquad A_{12}.\]The quench and steady-state discussion in Part III explains how these coefficients are simplified in practice. Part IV explains why $A_{12}$ can increase mutual information without guaranteeing synchronization. This section adds the operational cost:
\[A_{12}\text{ also increases circuit depth.}\]So the same cross-correlation structure appears in three roles:
| Coefficient | Role |
|---|---|
| $A_{12}$ in wavefunction | mixed Gaussian term |
| $A_{12}$ in correlations | source of shared information |
| $A_{12}$ in complexity | entangling preparation cost |
14. Compact Picture
The section can be summarized as:
| Step | Meaning |
|---|---|
| choose reference state | simple uncorrelated Gaussian |
| choose target state | time-evolved correlated Gaussian |
| choose Gaussian gate set | local scaling plus nonlocal entangling |
| compute circuit depth | compare reference and target quadratic forms |
| weak coupling | mostly local scaling |
| strong coupling | entangling cost grows |
| field dominated | logarithmic scaling with $\omega_c/\omega_R$ |
Thus circuit depth measures the operational price of preparing the correlated oscillator state. Mutual information may say that correlations are present, but circuit depth asks how expensive those correlations are to create.
Discussion