Compatibility and orientation are separate questions
An orthogonal complex structure must satisfy J²=−I and J*=−J. To be compatible with a specified symmetry representation V(g), it must also commute with every retained V(g). This set of compatible operators is an operator-valued space of possible phases; there need not be a preferred element before dynamics and orientation are supplied. [P02] [P03]
Now consider a skew-adjoint generator A. Its real flow exp(sA) preserves the inner product. The positive factor |A| describes the magnitude of the generator, while the polar factor gives its orientation on the nonstationary sector. In finite dimensions with nonzero frequencies, the relation is especially transparent.
The laboratory uses two independent real planes. Their signed rates are ω₁=1 and an adjustable ω₂. The choice of two planes and a block-diagonal generator is an illustration of the published construction. It does not classify every possible representation.
Try both signs, then a zero mode
Choose a candidate sign on each plane and change the second rate. The diagnostic −JA reports whether the candidate agrees with positive polar orientation. With ω₂ negative, the second plane requires −J₀. Set ω₂ to zero to see that neither sign is selected by its stationary flow.
Explore · Exact finite illustration
Change the inputs and compare the numerical readout. Playback begins only when you press Play.
The interactive values require JavaScript. The diagram and worked equations provide the same mathematical construction.
The animation shows the specified generator, regardless of whether the candidate phase passes the diagnostic. A rejected candidate therefore does not reverse the animation; it changes the attempted interpretation of that same dynamics.
The polar factor and the kernel
The second formula applies only to nonzero blocks. If a block vanishes, A supplies no orientation there. An extension of J to the kernel requires separate compatible data; an odd-dimensional real kernel cannot carry a complex structure by itself.
For a candidate J=diag(s₁J₀,s₂J₀), where each s is ±1, the diagonal factors of −JA are s₁ω₁ and s₂ω₂. A nonzero block must have a positive factor. A zero factor on the kernel leaves the candidate undecided. At repeated frequencies, other compatible structures may exist; this lab intentionally displays only the block-sign candidates.
What the finite illustration leaves visible
In infinite dimensions, A|A|⁻¹ is not always an everywhere-defined bounded formula. Polar decomposition and spectral calculus define the selected phase without requiring a positive spectral gap above zero, while operator-domain conditions remain essential. The dynamic-sector restriction is therefore more precise than simply dividing a matrix by its frequency. [P03] [B22]
Time reversal also has a clear planar meaning. The reflection K(x,y)=(x,−y) satisfies KJ₀=−J₀K. It is real linear but becomes antilinear after the complex structure is chosen. This example explains the change of scalar action without invoking a second observable time coordinate. [A03]
Continue to relative phase to ask which phase differences affect a completed measurement.

