Serially Conjoined Nuller System

• Serially Conjoined Nuller System is a multi-stage high-contrast imaging architecture that suppresses on-axis starlight by cascading a Lyot-type coronagraph with a fiber nuller.
• The first stage sculpts the residual stellar leak into a controlled mode, while the second stage uses modal filtering to achieve near-design contrast at off-design wavelengths.
• Experimental results demonstrate a factor of 20 improvement in off-design performance, offering a promising route for exoplanet imaging and advanced interferometry.

Searching arXiv for the specified paper and closely related nulling/coronagraph papers. A serially conjoined nuller system is a multi-stage high-contrast architecture in which one nuller is placed after another so that the first stage suppresses most on-axis starlight and shapes the residual leakage into a simple spatial mode, while the second stage is optimized to reject that remaining mode. In the most explicit recent realization, the first stage is a Lyot-type coronagraph, specifically the one-dimensional diffraction-limited coronagraph (1DDLC), and the second stage is a fiber nuller implemented with a Lyot-plane phase mask, relay optics, and a single-mode fiber (SMF). The experimental validation of this concept showed that serial coupling can recover near-design-wavelength contrast at an off-design wavelength where the 1DDLC alone is strongly degraded, thereby demonstrating a route to broader spectral robustness at small inner working angle (IWA) (Itoh et al., 24 Jul 2025).

1. Definition and governing idea

In high-contrast imaging, a nuller is an optical subsystem that suppresses the on-axis stellar field by destructive interference or spatial filtering while transmitting off-axis planetary light. A serially conjoined nuller system places two different nullers one after another in the optical train and co-designs them so that the first nuller controls the modal structure of the residual stellar leak and the second nuller rejects precisely that mode. In the 1DDLC-plus-fiber implementation, the first stage is a one-dimensional diffraction-limited coronagraph with promising features as a binary nuller and with small IWAs, but with strong sensitivity to spectral bandwidth and tilt aberrations; the second stage is a parity-based fiber nuller that uses a Lyot-plane mask to force the leak into a mode that cannot couple into an on-axis SMF (Itoh et al., 30 Jun 2026).

The central enabling property is the character of the 1DDLC’s off-design leakage. At the design wavelength $\lambda_c$, a properly designed 1DDLC yields perfect on-axis extinction within the Lyot stop in the ideal theory. For $\lambda \neq \lambda_c$, however, the cancellation is no longer exact, and a residual stellar leak appears. The decisive result is that this chromatic leak remains spatially simple at the Lyot plane: it has flat phase and the same complex amplitude profile as an on-axis pupil, up to a multiplicative scalar coefficient,

$$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$

Because the leak is “clean” in this sense, a downstream nuller can be designed once to suppress that pupil mode over a range of wavelengths, rather than having to cancel a wavelength-dependent aberrated field (Itoh et al., 24 Jul 2025).

This point is often misunderstood. Serial conjunction is not merely the stacking of two arbitrary nullers. Its logic depends on the first stage transforming chromatic or low-order leakage into a known, controlled mode at the Lyot plane. The second stage then functions as a mode-selective nuller rather than as a generic second coronagraph.

2. Optical architecture and modal mechanism

The experimentally validated architecture is: Entrance pupil $\rightarrow$ 1DDLC focal-plane mask $\rightarrow$ Lyot plane + Lyot stop $\rightarrow$ Lyot-plane binary phase mask (HWP+LP) $\rightarrow$ 1/100× relay $\rightarrow$ SMF $\rightarrow$ detector (Itoh et al., 24 Jul 2025).

The 1DDLC is a Lyot coronagraph whose focal-plane mask is one-dimensional and binary in amplitude and phase. Its mask function $M(x)$ takes real values in $\lambda \neq \lambda_c$0, where $\lambda \neq \lambda_c$1, so the mask is a complex-amplitude filter constrained to real values. In the related formulation,

$\lambda \neq \lambda_c$2

with $\lambda \neq \lambda_c$3 normalized by $\lambda \neq \lambda_c$4 (Itoh et al., 30 Jun 2026). At $\lambda \neq \lambda_c$5, the focal-plane mask redistributes the stellar PSF into high spatial frequencies outside the Lyot stop. The Lyot-plane field is

$\lambda \neq \lambda_c$6

and an appropriate $\lambda \neq \lambda_c$7 ensures

$\lambda \neq \lambda_c$8

Because the mask is one-dimensional, the coronagraph is inherently a binary-star nuller in the sense that it acts strongly along one dimension, and its off-axis throughput peaks near $\lambda \neq \lambda_c$9 (Itoh et al., 24 Jul 2025).

The second stage is a fiber nuller operating in the Lyot plane. In the experimental realization, the Lyot-plane mask is an achromatic phase mask built from a patterned half-wave plate plus a linear polarizer. It takes amplitude values $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$0 and $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$1 in a rectangular-wave pattern with period equal to the pupil width in the 1D nulling direction. A related description names this stage the Parity Fiber Nuller (PFN) and writes a two-dimensional Lyot-plane phase mask as

$$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$2

In both descriptions, the operational goal is the same: transform the clean stellar leak into a superposition that is odd in at least one coordinate (Itoh et al., 24 Jul 2025).

The SMF supplies the modal filtering. Its fundamental mode is even about the fiber center, so the coupling efficiency is governed by the overlap integral

$$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$3

If the focal-plane field is odd in one coordinate and the fiber mode is even, the integral vanishes. Since free-space propagation from pupil to focus is a Fourier transform and the Fourier transform preserves parity, an odd field in the Lyot plane remains odd at the fiber-injection focal plane. The on-axis leak is therefore rejected, while off-axis planetary light, which is not purely odd, retains non-zero coupling (Itoh et al., 24 Jul 2025).

The relay optics in the experimental system use two achromatic lenses, $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$4 and $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$5, to form a 1/100 demagnifying relay from Lyot plane to fiber focal plane. The scale is chosen so that $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$6 corresponds to $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$7 in the fiber injection focal plane, giving a $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$8 central lobe, matched to a Thorlabs P1-S630Y-FC-2 fiber with nominal mode-field diameter $$E_{\mathrm{Lyot}}(x,y;\lambda) \approx C(\lambda)\,P(x,y).$$9 at $\rightarrow$0 (Itoh et al., 24 Jul 2025).

3. Experimental demonstration and measured performance

The principal laboratory validation established that serial conjunction recovers strong contrast suppression at an off-design wavelength where the 1DDLC alone suffers a substantial chromatic leak. For the 1DDLC alone, earlier work had demonstrated contrast mitigation around $\rightarrow$1 at the design wavelength $\rightarrow$2. At $\rightarrow$3, about 6% shorter than $\rightarrow$4, the 1DDLC alone yielded only $\rightarrow$5 suppression. With the serially conjoined fiber nuller added, the measured contrast at that same off-design wavelength was

$\rightarrow$6

an improvement by a factor of $\rightarrow$7 relative to the 1DDLC alone at that wavelength (Itoh et al., 24 Jul 2025).

At the design wavelength, the measured minimum normalized contrast for the combined system was

$\rightarrow$8

slightly better than the previously demonstrated $\rightarrow$9 contrast for the 1DDLC alone, although the experiment was not primarily intended to improve the design-wavelength case. The off-design result is the more important one: the combined system approximately reaches the 1DDLC’s design-center contrast-mitigation ability demonstrated previously, suggesting that serial coupling compensates the coronagraph’s strong chromatic sensitivity at the $\rightarrow$0 level (Itoh et al., 24 Jul 2025).

Planet throughput remains finite at small separations. Simulations for the combined system show throughput peaks around

$\rightarrow$1

and for the actual commercial fiber the throughput map peaks around

$\rightarrow$2

with simulated throughput about $\rightarrow$3. Earlier measurements quoted in the source description gave measured off-axis throughput $\rightarrow$4 at that location, in agreement with simulation, while the ideal SMF and ideal mask design yielded a theoretical maximum coupling of $\rightarrow$5 for off-axis sources at diffraction-limited separation (Itoh et al., 24 Jul 2025).

These measurements support a specific bandwidth claim. For a raw-contrast goal around $\rightarrow$6, the effective bandwidth goes from narrow, stated as $\rightarrow$7–$\rightarrow$8 for the 1DDLC alone, to at least $\rightarrow$9 on the blue side when the fiber nuller is added. The inference is straightforward: because the leak mode shape is stable with wavelength and only its amplitude changes, the second-stage nuller is no longer limited by the same chromatic mismatch that constrains the first stage (Itoh et al., 24 Jul 2025).

A closely related account reports the off-design experiment as performed at a wavelength 6.5% longer than the design-center wavelength rather than 6% shorter, while retaining the same reported contrast mitigation of $\rightarrow$0 and the same factor-of-20 improvement relative to the 1DDLC alone. This discrepancy indicates that the precise sign and magnitude of the wavelength offset should be checked against the original experimental configuration, but it does not alter the reported demonstration that serial coupling suppresses the off-design leak by about one order of magnitude and a half (Itoh et al., 30 Jun 2026).

4. Relation to other serial nulling architectures

The serially conjoined nuller concept is not restricted to a single-aperture Lyot coronagraph followed by an SMF. In multi-aperture interferometry, the kernel-nuller at the VLTI is explicitly a two-stage serially conjoined nuller for four telescopes. Its first stage is a 4×4 beam combiner that generates one bright output and three dark outputs. Its second stage is a 3×6 combiner that performs all six pairwise combinations of the three dark outputs, applying a $\rightarrow$1 phase shift to create six nulled outputs with asymmetric transmission maps. A subsequent kernel matrix,

$\rightarrow$2

forms three kernel outputs as paired differences of those six nulled outputs. This architecture is serial both optically and in the measurement domain, and its kernel observables are designed to suppress certain instrumental leakage terms analogously to closure-phase and kernel-phase methods (Chingaipe et al., 2023).

The integrated-optics four-telescope Angel–Woolf nuller provides another concrete serial architecture. It uses three 2×2 directional couplers arranged in two stages: stage 1 performs pairwise nulling of two beam pairs, and stage 2 recombines the already-nulled outputs. In the ideal design, the four-aperture geometry and weighting yield a sixth-order angular null, $\rightarrow$3, near the broad minimum. Experimentally, the measured deep-nulling output followed a $\rightarrow$4 power law for delays larger than $\rightarrow$5 rad, while smaller detunings were limited by an experimental floor around 1:100. This result is significant for serial nulling because it shows that cascaded nulling stages can increase null order by cancelling lower-order terms in the stellar leakage expansion (Errmann et al., 2014).

Across these examples, serial conjunction serves two related purposes. In the single-aperture 1DDLC+fiber system, it is primarily a mode-engineering and mode-rejection strategy: the first stage sculpts the residual leak and the second stage filters it. In the kernel-nuller and Angel–Woolf cases, it is a hierarchical recombination strategy: the first stage produces dark outputs and the second stage further combines them to generate higher-order suppression or robust observables. The common thread is that downstream stages operate on structured outputs of upstream nullers rather than on the raw telescope field.

5. Robustness, limitations, and common misconceptions

The strongest robustness claim presently demonstrated is limited to contrast of order $\rightarrow$6. The combined 1DDLC-plus-fiber system showed $\rightarrow$7 at $\rightarrow$8 and $\rightarrow$9 at an offset wavelength, but future work is explicitly required to demonstrate the anticipated robustness for contrast-mitigation levels lower than about $\rightarrow$0, including the $\rightarrow$1–$\rightarrow$2 regime relevant to more demanding exoplanet applications (Itoh et al., 24 Jul 2025).

The principal experimental limitations are mechanical, thermal, detector-related, and modal. Dark-subtraction noise in the CCD is dominant in the fine-resolution scans because of temporal variations in dark current induced by detector temperature fluctuations. Thermal drifts are measurable: the piezo stage closed-loop drift over 10 hours is reported as peak-to-valley $\rightarrow$3, and focal-plane-mask adjustment drifts with temperature are treated as equivalent to tilt or pointing error. The SMF is non-ideal, with only a nominal mode-field specification, and Lyot-plane mask imperfections, polarization leakage, and residual optical aberrations can all reduce null depth and throughput (Itoh et al., 24 Jul 2025).

Alignment tolerances are correspondingly severe. The SMF rejects odd modes only if it is centered at the symmetry point of the field, and a mis-centering by $\rightarrow$4, about $\rightarrow$5 in the focal plane, can appreciably degrade null depth. To manage this, long scans are performed over a grid and the best point is selected in post-processing, with Gaussian filtering applied to the scan image to mitigate noise (Itoh et al., 24 Jul 2025).

A second common misconception is that the present serial system has already demonstrated broadband or aberration robustness in a general sense. The experimental evidence is narrower. It demonstrates bandwidth extension of at least 6% on one side of the design wavelength for a raw-contrast level of order $\rightarrow$6. The paper further states that the Lyot-plane mask is explicitly optimized to change the 1DDLC’s stellar leak due to wavelength deviation and the least-order tilt aberration into an odd function for at least one variable on the Lyot-stop plane, which suggests first-order mitigation of tilt sensitivity. However, no explicit experimental tilt-sensitivity measurements are given in that study (Itoh et al., 24 Jul 2025).

6. Significance for exoplanet imaging and possible extensions

For exoplanet imaging, the serially conjoined nuller is attractive because it combines small IWA with modal filtering. In the reported single-aperture implementation, throughput peaks near $\rightarrow$7 and remains at the few-percent level, specifically 5–10% near $\rightarrow$8, while raw stellar suppression reaches $\rightarrow$9. This places the concept in a regime where diffraction-limited separations are accessible without abandoning a fiber-fed architecture (Itoh et al., 24 Jul 2025).

The broader implication is a design philosophy of mode engineering. A front-end coronagraph is used not only to suppress the star directly but also to transform residual leakage into a small set of known spatial modes. A downstream fiber nuller, photonic device, or additional nulling stage can then be designed to remove those modes. The 1DDLC+fiber demonstration explicitly motivates extensions to improved Lyot-plane masks, optimized fiber mode-field diameters, multi-stage nulling, more general photonic implementations such as PIC-based systems, and a transition from 1D geometry to 2D masks for general high-contrast imaging (Itoh et al., 24 Jul 2025).

In interferometric variants, the same philosophy appears in different mathematical form. The VLTI kernel-nuller combines multi-stage optical nulling with kernel observables and Earth-rotation diversity, yielding effective IWA around 4 mas, uniformly covered kernel throughput maps over a $\rightarrow$0 h average, and astrometric recovery with scatter of only a few mas even with residual piston RMS as high as 150 nm. This suggests that serial conjunction can be paired with time diversity and robust observables to extend nulling performance beyond what a single optical null output provides (Chingaipe et al., 2023).

A plausible implication is that future serially conjoined nuller systems will be judged less by whether they contain two stages in series and more by whether the stages are mutually adapted: first-stage shaping of leakage, second-stage rejection of that specific leak, and preservation of useful off-axis throughput. In that stricter sense, the experimentally validated 1DDLC-plus-fiber configuration is a canonical example of a serially conjoined nuller system, while the kernel-nuller and integrated-optics multi-stage nullers demonstrate parallel realizations of the same systems principle.