betaFeedback

WorksPhysics and Astronomy

Investigating the Viability of Dedicated Microhertz Gravitational Wave Pulsar Timing

Abstract

The microhertz gravitational wave (GW) frequency regime remains one of the least explored regions of the GW spectrum. This is due to the physical limitations of our current and upcoming GW observatories such as LIGO and LISA, which are tuned to dHz and mHz respectively. Proposed experiments targeting the μμHz regime are ambitious, expensive and unlikely to be operational within the next 20 years. Pulsar Timing Arrays (PTAs), with their Galaxy-sized baselines, occupy the other end of the spectrum; the frequencies they are sensitive to are a function of the observing cadence and total dataset timespan. Existing PTA campaigns have cadences and timespans corresponding to the nanohertz regime, however the possibility exists that they can be 'pushed' to detect higher frequencies. This paper investigates the viability of a custom microhertz-sensitive PTA, observing a small number of millisecond pulsars with a daily cadence, over the course of 1 and 2 years respectively. We assume the use of both 100-m and 500-m class radio telescopes for these observations, and an idealised noise model with negligible systematics to determine the most optimistic detection threshold. We compute the sensitivity of our proposed high-cadence PTA configurations to power-law signals and individual GW sources. We find that a 500-m class telescope would likely be able to set new upper limits on individual sources and GW background signals in the 0.1−1 μ0.1-1~μHz range; while a 100-m class telescope would, in the best case, only match that of the performance of the EPTA second data release in the same range. Finally we consider a possible μμHz PTA consisting of SKA-Mid and DSA observing 24 pulsars; we find this performs on par with the 500-m class while using far less telescope time, presenting a more feasible solution to microhertz-GW pulsar timing.

Introduction

Pulsars are rapidly rotating neutron stars, that generally emit radiation across the radio spectrum, and appear, observationally, as pulsing objects, one pulse per rotation, as seen by radio telescopes on Earth [64]. The fastest-spinning sub-group of these objects are the millisecond pulsars (MSPs); these have been ‘spun up’ to rotational periods of the order of milliseconds through angular momentum transfer interactions with binary partners (e.g. [6] (Backer et al. 1982)). MSPs have extremely stable pulse profiles and timing properties [42, 19] and therefore can be used as highly accurate clocks, rivalling the precision of some of the best atomic clocks on Earth [29]. Their exceptional stability allows for precise timing measurements. Observations with an ensemble of pulsars can be used as a gravitational wave (GW) detector [60, 13], with the Earth and pulsar forming either end of a detection arm sensitive to very long-wavelength GWs. Timing is done by timestamping detected pulses, forming a collection of Times of Arrival (ToAs) for a given pulsar; observed ToAs are then subtracted from the corresponding predicted ToAs using a pulsar timing model, forming the so-called ‘residuals’ [13]. After accounting for the star’s rotational evolution, binary features, propagation effects etc., remaining non-zero residuals can be inferred as having been caused by passing GWs between Earth and the pulsar. Multiple MSPs can be used to form a combined detector with many Galaxy-sized arms, known as a Pulsar Timing Array (PTA; [17, 25]).

The minimum and maximum GW frequency of a PTA can be calculated from two parameters, namely the total observation time TobsT_{\mathrm{obs}} and the total number of ToAs NN [33, 28], as:

fmin⁡≈1Tobs,(1)f_{\min} \approx\frac{1}{T_{\mathrm{obs}}}, \tag*{(1)}
fmax⁡≈fNyq=N2Tobs,(2)f_{\max} \approx f_{\mathrm{Nyq}} = \frac{N}{2T_{\mathrm{obs}}}, \tag*{(2)}

where fNyqf_{\mathrm{Nyq}} is the Nyquist frequency. This however assumes uniform ToA sampling; PTAs unequally sample ToAs, given real-world telescope operational constraints, which in turn contributes to a higher upper frequency limit [38]. Notwithstanding, we can use (1) and (2) as reasonable estimates of the frequency range wherein a pulsar timing experiment is sensitive to GWs. PTAs generally have data sets spanning decades, observing with a weekly to monthly cadence [2, 72] (EPTA Collaboration et al. 2023), so correspondingly they are sensitive to GWs with frequencies in the range of 10−910^{-9} − 10−710^{-7} Hz. Meanwhile, Earth-based detectors, such as LIGO, operate in the range 1010 − 10310^{3} Hz [41]. By increasing the cadence of PTA observations to a daily rate, one can push the upper frequency limit into µHz frequencies.

The motivation for doing this is two-fold. Firstly, current PTA sensitivity is optimised to a frequency range where there is a degeneracy between possible contributions to a Gravitational Wave Background (GWB): from improperly-modelled Super-Massive Black Hole Binaries (SMBHBs) and from cosmological sources (also called ‘New Physics’) [2, 14]. While this degeneracy will likely be broken with further improvements to the existing PTA collaborations and methods (e.g. Goncharov et al. 2026), a potential concurrent solution would be to probe the microhertz GW frequency range, where SMBHBs are predicted not to radiate with the same strength as in the nanohertz range. There also exists microhertz-specific signals of interest, such as the gravitational wave signature of the QCD phase transition [58], which itself is a strong motivator to explore this regime. Secondly, there has been very limited exploration of the μ\muHz regime, owing to limitations in our detectors (see Figure 1), meaning it is potentially an as yet untapped discovery space. The Laser Interferometer Space Array (LISA) will eventually begin to fill in the gap from the millihertz side; with a planned launch date of 2035 [3]. Similarly, there are proposed space-based interferometers targetting the microhertz range specifically, such as μ\muAres, with planned operation in the 2040s [62], and another which proposes the use of asteroids as GW test masses [16]. Proposed next generation Earth-based GW observatories include the Einstein Telescope [55] and Cosmic Explorer [57] and target frequencies of Hz to kHz, with operation dates after 2035. There is an opportunity to attempt pulsar timing experiments in unexplored GW regimes between now and when these observatories come online, and to also provide some early insight into the microhertz regime, which could inform the designs of these purpose-built telescopes, each scientific megastructures in their own right. This paper explores the performance of a purpose-designed high-cadence PTA experiment exploring the μ\muHz GW frequency regime, compared to its existing counterparts, and if it would be theoretically capable of detecting and/or constraining various cosmological GW background signals. We also explore its ability to detect individual sources of GWs. It is organised as follows: in §2 we recap the potential sources of GWs that we expect in this frequency regime and outline the hypothetical observation campaign; in §3 we discuss the formalism used to compute sensitivity curves and pulsar noise; §4 presents and discusses the power-law sensitivity curves and sky-map sensitivity of a hypothetical μ\muHz-tuned PTA. For the remainder of this paper, natural units will be used, i.e. kB=c=ℏ=1k_{\mathrm{B}} = c = \hbar= 1, such that temperature and mass are expressed in terms of energy units. We take the Hubble constant to be H0=67.66 km s−1 Mpc−1H_0 = 67.66\ \mathrm{km\ s^{-1}\ Mpc^{-1}}, from Planck 2018 [54], also given in the dimensionless form h100=H0/100 km s−1 Mpc−1=0.6766h_{100} = H_0/100\ \mathrm{km\ s^{-1}\ Mpc^{-1}} = 0.6766.

Figure 1. The sensitivity curves of various current (EPTA DR2, aLIGO), upcoming (LISA, IPTA DR3, Einstein Telescope/ET) and proposed (Cosmic Explorer/CE, Big Bang Observer/BBO, µAres) GW detectors to a stochastic GWB in units of energy density. Note the current gap in the frequency coverage at the µHz regime. Figure generated from an online tool created by Mingarelli [45], see here.

Background

GW Sources

A gravitational wave background is predicted to be generated from the population of supermassive black hole binaries, peaking in the nHz frequency regime. The population of SMBHBs emitting μ\muHz GWs is expected to be drastically less than those emitting in the nHz regime, due to the fact that the time a SMBHB spends in a GW frequency bin, τres\tau_{\mathrm{res}}, is proportional to fgw−1f_{\mathrm{gw}}^{-1} [52]. This in turn means no stochastic background is expected from SMBHB in the μ\muHz region [1, 61], as only a handful of SMBHBs would populate a given frequency bin in this regime and would be measurably transient on PTA timescales [46]. Without a stochastic GW background due to SMBHB, any detection could be confidently attributed to alternative cosmological sources, demonstrating their existence and providing valuable insight into how the early Universe may have formed. The lack of a stochastic background of SMBHB binaries could also make detecting individual bright binaries more tractable.

First Order Phase Transitions

First order phase transitions (FOPTs) are predicted to have occurred during the early evolution of our Universe, as various fields decoupled from one-another [68, 30]. ‘First-order’ in this case indicates that transitions would have begun at random isolated points and nucleated at relativistic speeds outwards as spherical bubbles. While the growing spherical bubbles would not have generated GWs, the interaction of bubbles with each other and the turbulence generated in the surrounding plasma would have produced a GW signature. The frequency of this signature is proportional to the energy scale of the transition T∗T_\ast [9], and the relation is given as

fpeakFOPT∼10−6 Hz⋅T∗200 MeV(3)f_{\mathrm{peak}}^{\mathrm{FOPT}} \sim10^{-6}\ \mathrm{Hz}\cdot\frac{T_\ast}{200\ \mathrm{MeV}} \tag*{(3)}

The QCD phase transition is the predicted moment where the quark-gluon plasma cooled sufficiently to coalesce into hadrons. The temperature of this transition is calculated to be on the order of 200 MeV [53, 35]. This corresponds to a time on the order of 10 μ\mus after the Big Bang in standard cosmology, which places it after the Electroweak transition but before Big Bang nucleosynthesis [39]. The peak frequency of the QCD transition is approximately 1μ\muHz according to (3), making it a target of a μ\muHz search.

Cosmic Strings

Related to FOPTs are cosmic strings, which are 1-dimensional topological defects in spacetime that should arise at the interface between phase transition bubbles [49, 37]. They are emitters of GWs, and it is predicted that a network of cosmic strings would contribute to the GWB [7] (Damour & Vilenkin 2001). The main parameter determining the strength of a cosmic string background is the dimensionless cosmic string tension GμG\mu. Quelquejay Leclere et al. (2023) constrains this to Gμ≲10−9.9G\mu\lesssim10^{-9.9} for a particular cosmic string model, using a six-pulsar subset of the European PTA (EPTA) Data Release 2 (EPTA Collaboration et al. 2023). The correct model for a cosmic string GWB remains an open question, and it is uncertain if a power law is sufficient to capture its signature, or whether something like a broken power law would be more accurate [10]. We assume our desired target is a power law for reasons of simplicity, taking advantage of the boost in sensitivity this yields, explained in §3.1.

Domain Walls

Domain Walls are another topological defect that may have been produced by the aforementioned phase transitions in the early Universe and are a two-dimensional analogue to Cosmic Strings [37]. If they indeed formed they must have become unstable and annihilated themselves once the Universe cooled to a temperature TannT_{\mathrm{ann}} [71], and it is this self-annihilation that would produce a detectable GW signature, dependant on the value of TannT_{\mathrm{ann}}. The other parameter determining their emission strength is their surface tension σDW\sigma_{\mathrm{DW}} (see Saikawa 2017 for a review of domain walls). We follow the description of domain walls given in Saikawa (2017), which give the peak frequency and energy density of their emission spectrum to be

fpeakDW≈1.1×10−7 Hz⋅(g∗e10)1/2(g∗s10)−1/3(Tann1 GeV)(4)f_{\mathrm{peak}}^{\mathrm{DW}} \approx1.1 \times10^{-7}\ \mathrm{Hz} \cdot\left(\frac{g_{*e}}{10}\right)^{1/2}\left(\frac{g_{*s}}{10}\right)^{-1/3}\left(\frac{T_{\mathrm{ann}}}{1\ \mathrm{GeV}}\right) \tag*{(4)}
Ωgwpeak≈10−26h1002⋅(g∗e10)−4/3(σDW1 TeV3)2(Tann1 GeV)−4(5)\Omega_{\mathrm{gw}}^{\mathrm{peak}} \approx\frac{10^{-26}}{h_{100}^{2}} \cdot\left(\frac{g_{*e}}{10}\right)^{-4/3}\left(\frac{\sigma_{\mathrm{DW}}}{1\ \mathrm{TeV}^{3}}\right)^{2}\left(\frac{T_{\mathrm{ann}}}{1\ \mathrm{GeV}}\right)^{-4} \tag*{(5)}

Where g∗eg_{*e} and g∗sg_{*s} are the relativistic energy and entropy degrees of freedom, both functions of temperature, evaluated at TannT_{\mathrm{ann}}, and h100h_{100} is the dimensionless Hubble constant. We consider an annihilation temperature of Tann=1 GeVT_{\mathrm{ann}} = 1\ \mathrm{GeV}, corresponding to a time of 1 μs1\ \mathrm{\mu s} after the Big Bang [39]. We choose this as it corresponds to g∗e≈10≈g∗sg_{*e} \approx10 \approx g_{*s} [32], meaning fpeakDW∼10−7 Hzf_{\mathrm{peak}}^{\mathrm{DW}} \sim10^{-7}\ \mathrm{Hz}, thus making it a target for µHz GW pulsar timing.

Individual Sources

In addition to all-sky sensitivity, PTAs can also be ‘pointed’ in particular directions on the sky, so as to detect GWs from monochromatic individual sources (e.g. [14]); such as continuous GW emission from bright individual SMBHBs, or in the context of cosmological sources, GW bursts generated by cosmic strings [11]. We find the possibility of detecting individual SMBHB, also known as Continuous Gravitational Waves (CWs) a particularly interesting avenue of investigation in this regime, as there have already been searches of high-cadence subsets of PTA datasets for such targets, using both WSRT and FAST (e.g. [51, 8] (Yi et al. 2014; Zou et al. 2026)). We use the upper bound on the strain amplitude of optimal-location SMBHB, given by Zou et al. (2026), as a benchmark for the required sensitivity needed to begin probing individual GW sources in the microhertz regime. They give this as 8.9×10−148.9 \times10^{-14} at 1 μHz1\ \mathrm{\mu Hz}, which was established using a high-cadence subset of observations

of PSR J1909–3744 from the second IPTA Data Release.

Observation Campaign

We assume a Westerbork Synthesis Radio Telescope (WSRT)-like observatory for the sensitivity calculations undertaken in this paper, using the 92-cm receiver and all fourteen 25-m dishes. WSRT serves as a good representation for a 100-metre class radio telescope; furthermore it is potentially on the cusp of an upgrade so that investigating the µHz PTA question is also a feasibility study for a potential science driver for such an upgrade. We also repeated our analyses using Five-hundred-metre Aperture Spherical Telescope (FAST)-like parameters, which has approximately 1 order of magnitude better gain [48]. The parameters we use for each of these telescopes are included in Table 1. For the simulated PTA, we choose four “good timers”1 from the

Telescope Parameters
SymbolNameWSRT Val.FAST Val.
fobsf_{\mathrm{obs}}Observation Frequency340 MHz1250 MHz*
AeffA_{\mathrm{eff}}Effective Area2750 m2^{2}35340 m2^{2}
tintt_{\mathrm{int}}ToA Integration Time3600 s3600 s*
npoln_{\mathrm{pol}}# of Polarisations22
Δf\Delta fTelescope Bandwidth70 MHz500 MHz*
TsysT_{\mathrm{sys}}System Temperature250 K50 K*

Table 1. The parameters and corresponding values used in equation (16) for a WSRT- and FAST-like observatory. The effective area of WSRT is given for fourteen 25-m dishes with a 40% aperture efficiency. We take the FAST values from Jiang et al. [34] and Xu et al. (2025). The effective area of FAST is given for a 300-m illuminated aperture with 50% efficiency. Any row marked with an asterisk indicates the value was chosen by the authors.

EPTA DR22 list, see Table 2.

PulsarσradWSRT\sigma_{\rm rad}^{\rm WSRT} (ns)σradFAST\sigma_{\rm rad}^{\rm FAST} (ns)PP (ms)W50W_{50} (ms)Smean400S_{\rm mean}^{400} (mJy)Smean1400S_{\rm mean}^{1400} (mJy)
J1012+5307586265.260.2304
J1022+10015225816.450.5753.9
J1744−1134389164.070.1182.6
J1857+09433543835.360.5205.0

Table 2. The pulsar ensemble used in this paper and corresponding properties PP (pulsar period), W50W_{50} (pulse width at 50% of the maximum) and SmeanS_{\rm mean} (the mean flux density at 400 MHz, 1400 MHz), taken from the ATNF catalogue. The quantity σradWSRT\sigma_{\rm rad}^{\rm WSRT} is the radiometer error computed via Equation (16) for a WSRT-like telescope using Smean400S_{\rm mean}^{400}. We use the same pulsars for the FAST-like sensitivity curve, computing σradFAST\sigma_{\rm rad}^{\rm FAST} with the modified telescope parameters in Table 1 and Smean1400S_{\rm mean}^{1400}.

We present three observational campaigns: Campaign A, with a cadence of 2 ToAs per pulsar per day, assuming an hour-long observation to be integrated to form each ToA, corresponding to a daily 8 hours of observation, and a total observation baseline length of 1 year. Campaign B has a cadence of 4 ToAs per pulsar per day, with the same integration time, equating to a daily 16 hours of observation, with a total baseline of 2 years. All ToAs are assumed to be sampled at equal intervals. Campaigns A and B correspond approximately to a Nyquist upper frequency limit of fmax⁡≈10f_{\max} \approx10 µHz, and 20 µHz respectively. Both these campaigns, especially B, are extremely intensive (Campaign B requires 11,680 hours on WSRT/FAST), so we investigate a third configuration, Campaign C, involving the upcoming Square Kilometer Array (SKA; Shannon et al. [63]), specifically SKA-Mid, and the Deep Synoptic Array (DSA; Hallinan et al. [21]). These are both next-generation interferometric radio arrays with substantial time dedicated towards pulsar timing. We assume the DSA will observe 12 MSPs each at a 1 day cadence, and choose a radiometer noise σradDSA=50\sigma_{\rm rad}^{\rm DSA} = 50 ns, as an order-of-magnitude representation of the DSA3 (Hallinan et al. [21]). For simplicity we assume the SKA will also observe 12 MSPs with the same cadence; and we take σradSKA=50\sigma_{\rm rad}^{\rm SKA} = 50 ns. We assume this requires an integration time of 30 minutes for each ToA for both the DSA and SKA-Mid. In total this adds up to be 6 hours per day per telescope. We believe both these telescopes will be capable of simultaneous timing of multiple sources, and so consider the actual total observation time for this campaign to be less; assuming both of them use 4 dedicated tied-array beams simultaneously, this would correspond to a total 1.5 hour per day per telescope. We assume a total observation length of 2 years. We do not select 24 MSPs but rather generate random sky-positions for each as these same telescopes will themselves identify a substantial population of currently unknown MSPs. We believe this setup to be a feasible timing regime for these two observatories, and works hand-in-hand with their proposed timing projects, as these ToAs can be included in their nanohertz-targeting timing.

Methodology

Sensitivity Curves

In this section we follow the formalism derived by Thrane & Romano [66] to characterise the sensitivity of a PTA to a power-law signal. We provide a brief summary of their derivation below. The expected power signal-to-noise ratio (SNR) ρ\rho, for a cross-correlation search for an unpolarised and isotropic stochastic gravitational wave background is

ρ=2Tobs[∫fmin⁡fmax⁡∑i=1M∑j=1MΓij2(f)Sh2(f)Pni(f)Pnj(f) df]1/2,(6)\rho= \sqrt{2T_{\rm obs}}\left[\int_{f_{\min}}^{f_{\max}}\sum_{i=1}^{M}\sum_{j=1}^{M}\frac{\Gamma_{ij}^{2}(f)S_h^{2}(f)}{P_{n_i}(f)P_{n_j}(f)}\,df\right]^{1/2}, \tag*{(6)}

where MM is the number of GW detectors, in this case pulsars. For two given detectors ii, jj, Γij\Gamma_{ij} is the overlap reduction function between them and PniP_{n_i} and PnjP_{n_j} are the respective noise power spectral densities. The gravitational wave power spectral density from a target background is ShS_h. From this one can infer an effective strain noise power spectral density

Sn(f)≡[∑i=1M∑j=1MΓij2(f)Pni(f)Pnj(f)]−1/2,(7)S_n(f) \equiv\left[\sum_{i=1}^{M}\sum_{j=1}^{M}\frac{\Gamma_{ij}^{2}(f)}{P_{n_i}(f)P_{n_j}(f)}\right]^{-1/2}, \tag*{(7)}

and the corresponding noise characteristic strain hnh_n and noise energy density Ωn\Omega_n can be calculated using the following relations

hn(f)=fSn(f),(8)h_n(f) = \sqrt{fS_n(f)}, \tag*{(8)}
Ωn(f)=2π23H02f3Sn(f),(9)\Omega_n(f) = \frac{2\pi^2}{3H_0^2}f^3S_n(f), \tag*{(9)}

respectively.4 When the signal of interest follows a power law, one can integrate over frequency to account for the higher sensitivity of a PTA to power-law signals

Ωα=ρ2Tobs[∫fmin⁡fmax⁡(f/fref)2αΩn2(f) df]−1/2.(10)\Omega_\alpha= \frac{\rho}{\sqrt{2T_{\rm obs}}}\left[\int_{f_{\min}}^{f_{\max}}\frac{(f/f_{\rm ref})^{2\alpha}}{\Omega_n^2(f)}\,df\right]^{-1/2}. \tag*{(10)}

For a set of indices α\alpha, the power-law integrated sensitivity curve is then given as

Ωgw=max⁡α[Ωα(ffref)α].(11)\Omega_{\rm gw} = \max_\alpha\left[\Omega_\alpha\left(\frac{f}{f_{\rm ref}}\right)^\alpha\right]. \tag*{(11)}

This can be converted to other forms using (8), (9). For a PTA, the overlap reduction function is given as

Γij(f)=13(2πf)2ζij,(12)\Gamma_{ij}(f) = \frac{1}{3(2\pi f)^2}\zeta_{ij}, \tag*{(12)}

where

ζij≡32(1−cos⁡θij2)log⁡(1−cos⁡θij2)−14(1−cos⁡θij2)+12+12δij(13)\zeta_{ij} \equiv\frac{3}{2}\left(\frac{1-\cos\theta_{ij}}{2}\right)\log\left(\frac{1-\cos\theta_{ij}}{2}\right)-\frac{1}{4}\left(\frac{1-\cos\theta_{ij}}{2}\right)+\frac{1}{2}+\frac{1}{2}\delta_{ij} \tag*{(13)}

is the Hellings-Downs factor (Hellings & Downs [25]) such that it is normalised for a single pulsar (ζii=1\zeta_{ii}=1). When characterising PTA sensitivity to individual sources of GWs, we cannot use the formalism thus far developed, but refer the reader to Hazboun et al. [24], which details the method used. In short, we compute SnS_n for each sky position on the ecliptic and calculate the strain amplitude5^{5} h0(n^)h_0(\hat{n}) of a source at sky position n^\hat{n} emitting GWs at a frequency of f0f_0 that would yield a signal-to-noise ratio of ρ\rho, as

h0(n^)=ρSn(f0,−n^)Tobs.(14)h_0(\hat{n})=\rho\sqrt{\frac{S_n(f_0,-\hat{n})}{T_{\mathrm{obs}}}} . \tag*{(14)}

The library hasasia (Hazboun et al. [23]) was developed to compute sensitivity curves and sky maps of a given PTA without the need to use computationally expensive simulations. It implements the formalism detailed above. We use this library to produce data in Figure 2, Figure 3, Figure 4. We verified the power-law sensitivity output of hasasia by independently developing software to compute Equation (11). After successfully verifying the calculations we proceeded using hasasia due to its additional features and polish.

Figure 2. The theoretical minimum energy density sensitivity to a 5-SNR GW background signal of the μHz-tuned PTA, using both WSRT-like and FAST-like observatories (top: Campaign A; bottom: Campaign B). This is compared to simulated GWB signals generated by a QCD (T∗∼200T_{\ast} \sim200 MeV) phase transition, a cosmic string network with tension Gμ∼10−10G\mu\sim10^{-10} and domain walls with annihilation temperature of Tann∼1T_{\mathrm{ann}} \sim1 GeV. Also included as a benchmark is an idealised sensitivity of EPTA DR2, computed using the same method with the following parameters: 25 pulsars observed with 20 yr−1^{-1} cadence over 24 years with a uniform σrad=1\sigma_{\mathrm{rad}} = 1 μs.

Figure 3. Top: Sky map of the ecliptic sphere from the hypothetical four-pulsar timing array using a WSRT-like telescope for the two different campaigns; Campaign A (Tobs=1T_{\mathrm{obs}} = 1 yr, 2 ToAs/day) on the left and Campaign B (Tobs=2T_{\mathrm{obs}} = 2 yr, 4 ToAs/day) on the right. Bottom: The same as above but using a FAST-like telescope instead. All four sky maps display the faintest detectable strain amplitude hh of a 5-SNR signal at a GW frequency of 1 μ\muHz, for a given campaign. The optimal-direction value, hmin⁡h_{\min}, is also presented. The four pulsars are denoted by white stars.

Power-law sensitivity and ecliptic sky-map for the proposed DSA+SKA-Mid PTA

Figure 4. Left: The power-law sensitivity of the proposed DSA+SKA-Mid PTA (24 MSPs, 1 ToA per day), compared to FAST Campaign B. Also included are a selection of potential cosmological backgrounds. Right: An ecliptic sky-map given in the same units as Figure 3 using the proposed DSA+SKA-Mid PTA. The pulsars are randomly distributed across the sky.

Pulsar Timing Noise

Like all experimental apparatus, PTAs have a noise component that must be sufficiently modelled to achieve desired sensitivity. This noise can arise from the pulsar itself, the intermediate space between it and Earth, observatory/instrumentation/clock noise and errors arising from outdated ephemerides. Broadly speaking, PTA noise can be divided into two categories, time-uncorrelated “white” noise and time correlated “red” noise (Lentati et al. [40]).

For the purposes of this feasibility study we are only concerned with order of magnitude predictions of sensitivity, we assume a very simplified model of white noise; following Thrane & Romano [66], we take the power spectrum to be

PWN=2Δtσrad2,(15)P_{\mathrm{WN}}=2\Delta t\sigma_{\mathrm{rad}}^2 , \tag*{(15)}

where Δt=1/cadence\Delta t=1/\mathrm{cadence} and σrad\sigma_{\mathrm{rad}} is the radiometer error. For each pulsar we calculate σrad≈W/SNR\sigma_{\mathrm{rad}}\approx W/\mathrm{SNR} (Backer 1989), which we expand as

σrad≈β2kBTsysAeffSmeantintnpolΔfW3p−W,(16)\sigma_{\mathrm{rad}}\approx\beta\frac{2k_{\mathrm{B}}T_{\mathrm{sys}}}{A_{\mathrm{eff}}S_{\mathrm{mean}}\sqrt{t_{\mathrm{int}}n_{\mathrm{pol}}\Delta f}}\sqrt{\frac{W^3}{p-W}} , \tag*{(16)}

using the radiometer equation (Lorimer & Kramer [43]). The ‘correction factor’ β\beta accounts for various effects such as digitisation loss, pulse shape mismatch, etc., which we take as β=1.25\beta=1.25 as a catch-all. The remaining variables in (16) are given in Table 1 and Table 2, and the computed values for σrad\sigma_{\mathrm{rad}} are given in Table 2. We use pulsar properties taken from version 2.7.0 of the ATNF pulsar catalogue6 (Manchester et al. [44]).

Red noise is modelled as a power law (Lentati et al. [40]), i.e.

PRN∝ARN(f)−γRN,(17)P_{\mathrm{RN}}\propto A_{\mathrm{RN}}(f)^{-\gamma_{\mathrm{RN}}} , \tag*{(17)}

where ARNA_{\mathrm{RN}} and γRN\gamma_{\mathrm{RN}} are respectively the amplitude and spectral index of a given red noise process. Each red noise process will have a different value for each of these parameters, and one or more of these may couple with and muffle a GWB power-law signature. We omit red noise from our analysis and so we present best-case estimates. Depending on the pulsar, jitter is also a substantial factor to consider for a telescope as sensitive as FAST (Liu et al. [42]; Wang et al. [67]), but we likewise omit it for simplicity.

Results & Discussion

Power-Law Sensitivity

We include GW curves from the three cosmological sources previously mentioned (phase transitions, cosmic strings and domain walls) as possible GW background targets in this regime. We use simulated data for a cosmic string-generated stochastic GWB from Blanco-Pillado & Olum [7]. This simulation modelled the growth and scaling of a cosmic string network using the Nambu-Goto approximation, to compute its expected GW profile. We select a string tension of 10−1010^{-10}, rounding to the 10−9.910^{-9.9} upper limit given in Quelquejay Leclere et al. (2023). We also include the predicted GW signature of a QCD phase transition (corresponding to Tann≈200T_{\mathrm{ann}}\approx200 MeV) taken from Renzini et al. [58]. Finally we compute a domain wall background signal using results of numerical simulations (e.g. Hiramatsu et al. [26] (2014); Kawasaki & Saikawa [36]) which show that

Ωgw∝{f3for f<fpeakf−1for f>fpeak\Omega_{\mathrm{gw}}\propto \begin{cases} f^3 & \text{for } f<f_{\mathrm{peak}}\\ f^{-1} & \text{for } f>f_{\mathrm{peak}} \end{cases}

with fpeakf_{\mathrm{peak}} given by (4). We assume σDW1/3=106\sigma_{\mathrm{DW}}^{1/3}=10^6 GeV and Tann=1T_{\mathrm{ann}}=1 GeV, which are selected from the allowable region of the parameter space given in Saikawa (2017), and the peak energy density is then calculated using (5).

In Figure 2 the theoretical sensitivity of the μ\muHz-tuned PTA, for Campaign A and B, computed using hasasia, is compared to the projected GW energy density of these sources. We also include an EPTA curve as a benchmark for current PTAs (consisting of 25 pulsars observed with a cadence of 20 yr−1^{-1} for 24 years), which is generated using the same radiometer noise-only assumption. We use this simulated data rather than actual data to allow for a fair comparison with the proposed PTA campaigns, considering we omit non-radiometer noise from our PTA models. Considering the Campaign A scenario, the WSRT-like curve reaches a minimum Ωgw\Omega_{\mathrm{gw}} sensitivity of approximately 10−410^{-4}, occuring at a GW frequency of 0.1 μ\muHz. The FAST-like observatory reaches a sensitivity of ∼10−6\sim10^{-6} at the same frequency. While the WSRT-like curve remains well above the upper limits for all GWB sources, the FAST-like curve gets within an order of magnitude of the QCD phase transition signature. Considering the Campaign B scenario, we see a two-order-of-magnitude improvement to both WSRT and FAST curves. While this has no effect on a WSRT-PTA’s ability to constrain the presented cosmological backgrounds, it now performs better than the EPTA benchmark above 0.1 μ\muHz. The FAST-PTA curve is now almost 4 orders of magnitude better than the EPTA benchmark, and possessing the possibility of constraining both phase transition and domain wall signal. The frequency with best sensitivity has shifted lower, as expected from the larger observation campaign length. While it might be obvious to consider increasing TobsT_{\mathrm{obs}} as a means of improving sensitivity, increasing the total observation time to decadal lengths will lower the maximum detectable frequency to nanohertz values in accordance with Equation (2), which is counter to the goal of probing the microhertz regime. Finally we consider power-law sensitivity of the DSA+SKA-Mid proposal in the left-hand plot in Figure 4. Comparing it to FAST Campaign B we see that it has the same order of magnitude sensitivity (with Ωgwmin∼10−8\Omega_{\mathrm{gw}}^{\mathrm{min}}\sim10^{-8}), making it capable of probing both the phase transition and domain wall background as modelled in this paper. We expect the DSA+SKA-Mid PTA would perform better than FAST Campaign B had we included non-radiometer noise, as the factor of six more pulsars would allow for better noise modelling.

Individual Source Sensitivity

We take the strain amplitude upper-limit from individual SMBHBs of h=8.9×10−14h = 8.9 \times10^{-14} at 1 μHz from Zou et al. (2026) as our benchmark for a continuous GW signal in the microhertz regime. Using the effective noise power spectral density computed via hasasia and Equation (14), we compute the minimum GW strain amplitude hh that would have a signal-to-noise ratio of 5, using campaigns A and B on both the WSRT- and FAST-like telescopes (at 1 μHz). We compute hh for each sky position, which is shown in Figure 3. The optimal sky-location is predicted, for this set of pulsars, to be latitude +83∘+83^{\circ}, longitude 207∘207^{\circ} on the ecliptic (RA 17h3m, Dec +63∘+63^{\circ}); but as is clear from the colour gradient in Figure 3, the whole region 30∘30^{\circ} above the ecliptic plane shares the same order-of-magnitude sensitivity. Neither campaigns using a WSRT-like telescope would see a GW signal with the strain amplitude, given, as both predict minimum strain of >3×10−13> 3 \times10^{-13}. However a better result is seen in both FAST-like campaigns, with a predicted optimal-direction minimum strain amplitude of 3.47×10−143.47 \times10^{-14} for Campaign A, and 1.73×10−141.73 \times10^{-14} for Campaign B. Thus if there was such a target source in the optimal sky location, and it was emitting at the upper-limit strain amplitude established in Zou et al. (2026), both FAST campaigns would detect it in this scenario. Admittedly this is a compounding list of best-case scenarios, but as the sky-maps show the h∼10−14h \sim10^{-14} region actually covers a substantial fraction of the sky. At the very least it would likely establish new upper-limits on individual GW sources in the microhertz regime.

The right-hand plot of Figure 4 shows the minimum strain amplitude across sky of a 5-SNR signal detectable by the DSA+SKA-Mid campaign. As the plotted MSPs are just randomly chosen points on the sky, we will ignore the specific optimal-direction, as this will change with an actual choice of pulsars. The best strain amplitude detectable in this idealised scenario is 3.27×10−143.27 \times10^{-14}, which is on par with FAST Campaign A and a factor of about 3 times worse of FAST Campaign B. However like mentioned in §4.1, this would likely be different once other noise components are considered. A major advantage of this over just using the singular telescope is the uniform sky-coverage that can be achieved using the Northern-hemisphere DSA and the Southern-hemisphere SKA-Mid.

Conclusion

As calculated in §4, a 100-m class (WSRT-like) PTA would struggle to provide novel insight into the microhertz GW regime. The power-law sensitivity of Campaign A performs on par with the idealised EPTA at all frequencies >0.1> 0.1 μHz, and Campaign B is only an order of magnitude better, but remains well above any of the proposed GWB upper-limits. It also would not be able to detect an individual source emitting GWs at the upper-limit established in Zou et al. (2026). The 500-m class (FAST-like) PTA campaigns are both much more promising, with both campaigns having a better power-law sensitivity than the white-noise-only EPTA curve in the >0.1> 0.1 μHz range. Thus it is likely a FAST-like Campaign B would be able to set new upper limits on the cosmological backgrounds mentioned. Likewise it could detect an individual source at the upper-limit of 8.9×10−148.9 \times10^{-14} for Campaigns A and B respectively, and even probe deeper, meaning it could also establish new upper bounds for individual GW sources in the μHz regime. An important factor to consider however is the telescope time such a campaign would require. Over the course of 2 years, campaign B would require 11,680 hours of telescope time. Given the competition for telescope time on an instrument such as FAST, such an observational campaign is incredibly unlikely. Comparatively, a DSA and SKA-Mid μHz PTA would require approximately 1,095 hours for each telescope over the same two-year timespan, and would achieve similar results to the FAST Campaign B. Furthermore, the DSA has already committed to this order of magnitude pulsar timing cadence [21], and we believe SKA-Mid capable of it likewise; thus it is a matter of when, not if, timing data like these become available. It is important to reiterate that this is an idealised-noise study, and as such the results must be understood as representing the most optimistic performance of these proposed PTAs. Further work must be done to investigate the noise environment of the microhertz regime and its effect on the sensitivities of each campaign discussed above, in preparation for timing data from telescopes like the DSA and SKA.

Any ToAs generated in a microhertz-targeting campaign are reusable, as they can be combined with other datasets, like the International Pulsar Timing Array, meaning timing experiments like this have use outside of microhertz-specific searches. High-cadence subsets are of value to the PTA community as they have already been used to put novel constraints on individual sources of GWs (e.g. [8] (Zou et al. 2026). Despite these efforts, the microhertz regime still stands as one of the least-explored regions of the gravitational wave spectrum. There are clear science motivators to explore this regime, including the aforementioned cosmological sources and bright individual supermassive black hole binaries. Thus as we await the development of new GW detectors targeting microhertz GW frequencies (e.g. [62, 16]), the boundaries of pulsar timing should be pushed and tested to its limit.

Acknowledgements

We thank Gemma Janssen for supplying information on WSRT. We also thank Boris Goncharov for taking the time to review this manuscript and for providing valuable feedback. This work has emanated from research conducted with the financial support of Taighde Éireann - Research Ireland under Grant number GOIPG/2025/5806.

Code & Data Availability

The power-law sensitivity curves, sky sensitivity maps and all other related data, excluding the cosmological background data, are deposited at 10.5281/zenodo.23690555 [65]. All code written for this work is available at the GitHub repository cnsher/microhertz_PTA.

This research made use of the following libraries: NumPy [22], pandas (pandas development team 2020), Matplotlib [31], HealPy [73, 20], Hasasia (Hazboun et al. 2021), SciencePlots [18], AstroPy [4], BeautifulSoup and Requests.

This paper was built using the Open Journal of Astrophysics LaTeX template. The OJA is a journal which provides fast and easy peer review for new papers in the astro-ph section of the arXiv, making the reviewing process simpler for authors and referees alike. Learn more at http://astro.theoj.org.

References

  1. [1]Afzal, A., Agazie, G., Anumarlapudi, A., et al. 2023, ApJ, 951, L11, doi: 10.3847/2041-8213/acdc91DOI
  2. [2]Agazie, G., Alam, M. F., Anumarlapudi, A., et al. 2023, ApJ, 951, L9, doi: 10.3847/2041-8213/acda9aDOI
  3. [3]Amaro-Seoane, P., Audley, H., Babak, S., et al. 2017, arXiv e-prints, arXiv:1702.00786, doi: 10.48550/arXiv.1702.00786
  4. [4]Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167, doi: 10.3847/1538-4357/ac7c74
  5. [6]Bhattacharya, D., & van den Heuvel, E. P. J. 1991, Phys. Rep., 203, 1, doi: 10.1016/0370-1573(91)90064-SDOI
  6. [7]Blanco-Pillado, J. J., & Olum, K. D. 2017, Phys. Rev. D, 96, 104046, doi: 10.1103/PhysRevD.96.104046
  7. [8]Caballero, R. N., Xu, H., Lee, K., et al. 2025, Research in Astronomy and Astrophysics, 25, 035022, doi: 10.1088/1674-4527/adb5d6DOI
  8. [9]Caprini, C., Durrer, R., & Servant, G. 2008, Phys. Rev. D, 77, 124015, doi: 10.1103/PhysRevD.77.124015
  9. [10]Caprini, C., Figueroa, D. G., Flauger, R., et al. 2019, J. Cosmology Astropart. Phys., 2019, 017, doi: 10.1088/1475-7516/2019/11/017DOI
  10. [11]Damour, T., & Vilenkin, A. 2000, Phys. Rev. Lett., 85, 3761, doi: 10.1103/PhysRevLett.85.3761DOI
  11. [13]Detweiler, S. 1979, ApJ, 234, 1100, doi: 10.1086/157593DOI
  12. [14]EPTA Collaboration, Antoniadis, J., Babak, S., et al. 2024a, A&A, 685, A94, doi: 10.1051/0004-6361/202348433
  13. [16]Fedderke, M. A., Graham, P. W., & Rajendran, S. 2022, Phys. Rev. D, 105, 103018, doi: 10.1103/PhysRevD.105.103018DOI
  14. [17]Foster, R. S., & Backer, D. C. 1990, ApJ, 361, 300, doi: 10.1086/169195
  15. [18]Garrett, J. D. 2021, garrettj403/SciencePlots, 1.0.9, Zenodo, doi: 10.5281/zenodo.4106649
  16. [19]Ghosh, A., Bhattacharyya, B., Sharan, R., et al. 2025, ApJ, 991, 135, doi: 10.3847/1538-4357/adff5aDOI
  17. [20]Górski, K. M., Hivon, E., Banday, A. J., et al. 2005, ApJ, 622, 759, doi: 10.1086/427976
  18. [21]Hallinan, G., Ravi, V., Weinreb, S., et al. 2019, in Bulletin of the American Astronomical Society, Vol. 51 (AIP), 255, doi: 10.48550/arXiv.1907.07648DOI
  19. [22]Harris, C. R., Millman, K. J., van der Walt, S. J., et al. 2020, Nature, 585, 357, doi: 10.1038/s41586-020-2649-2DOI
  20. [23]Hazboun, J., Romano, J., & Smith, T. 2019a, Journal of Open Source Software, 4, 1775, doi: 10.21105/joss.01775DOI
  21. [24]Hazboun, J. S., Romano, J. D., & Smith, T. L. 2019b, Phys. Rev. D, 100, 104028, doi: 10.1103/PhysRevD.100.104028DOI
  22. [25]Hellings, R. W., & Downs, G. S. 1983, ApJ, 265, L39, doi: 10.1086/183954DOI
  23. [26]Hiramatsu, T., Kawasaki, M., & Saikawa, K. 2010, J. Cosmology Astropart. Phys., 2010, 032, doi: 10.1088/1475-7516/2010/05/032DOI
  24. [28]Hobbs, G., Jenet, F., Lee, K. J., et al. 2009, MNRAS, 394, 1945, doi: 10.1111/j.1365-2966.2009.14391.xDOI
  25. [29]Hobbs, G., Guo, L., Caballero, R. N., et al. 2020, MNRAS, 491, 5951, doi: 10.1093/mnras/stz3071DOI
  26. [30]Hogan, C. J. 1986, MNRAS, 218, 629, doi: 10.1093/mnras/218.4.629DOI
  27. [31]Hunter, J. D. 2007, Computing in Science & Engineering, 9, 90, doi: 10.1109/MCSE.2007.55DOI
  28. [32]Husdal, L. 2016, Galaxies, 4, 78, doi: 10.3390/galaxies4040078DOI
  29. [33]Jenet, F. A., Hobbs, G. B., Lee, K. J., & Manchester, R. N. 2005, ApJ, 625, L123, doi: 10.1086/431220DOI
  30. [34]Jiang, P., Yue, Y., Gan, H., et al. 2019, Science China Physics, Mechanics, and Astronomy, 62, 959502, doi: 10.1007/s11433-018-9376-1DOI
  31. [35]Kapusta, J., Muller, B., & Rafelski, J. 2003, Quark-Gluon Plasma: Theoretical Foundations An Annotated Reprint Collection (London, England: Elsevier Science)
  32. [36]Kawasaki, M., & Saikawa, K. 2011, J. Cosmology Astropart. Phys., 2011, 008, doi: 10.1088/1475-7516/2011/09/008DOI
  33. [37]Kibble, T. W. B. 1976, Journal of Physics A Mathematical General, 9, 1387, doi: 10.1088/0305-4470/9/8/029DOI
  34. [38]Koen, C. 2006, MNRAS, 371, 1390, doi: 10.1111/j.1365-2966.2006.10762.xDOI
  35. [39]Kolb, E. W., & Turner, M. S. 1990, The Early Universe, Vol. 69 (Addison-Wesley)
  36. [40]Lentati, L., Shannon, R. M., Coles, W. A., et al. 2016, MNRAS, 458, 2161, doi: 10.1093/mnras/stw395DOI
  37. [41]LIGO Scientific Collaboration, Aasi, J., Abbott, B. P., et al. 2015, Classical and Quantum Gravity, 32, 074001, doi: 10.1088/0264-9381/32/7/074001DOI
  38. [42]Liu, K., Keane, E. F., Lee, K. J., et al. 2012, MNRAS, 420, 361, doi: 10.1111/j.1365-2966.2011.20041.xDOI
  39. [43]Lorimer, D. R., & Kramer, M. 2012, Handbook of Pulsar Astronomy (Cambridge University Press)
  40. [44]Manchester, R. N., Hobbs, G. B., Teoh, A., & Hobbs, M. 2005, AJ, 129, 1993, doi: 10.1086/428488DOI
  41. [45]Mingarelli, C. M. F. 2026, Energetic Ceilings of Astrophysical Gravitational-Wave Backgrounds. https://arxiv.org/abs/2601.18859
  42. [46]Mingarelli, C. M. F., Casey-Clyde, J. A., Chang, Y. T., et al. 2026, arXiv e-prints, arXiv:2603.13643, doi: 10.48550/arXiv.2603.13643DOI
  43. [47]Moore, C. J., Cole, R. H., & Berry, C. P. L. 2014, Classical and Quantum Gravity, 32, 015014, doi: 10.1088/0264-9381/32/1/015014DOI
  44. [48]Nan, R. L., Din, J. C., et al. 2011, International Journal of Modern Physics D, 20, 989, doi: 10.1142/S0218271811019335
  45. [49]Nielsen, H. B., & Olesen, P. 1973, Nuclear Physics B, 61, 45, doi: 10.1016/0550-3213(73)90350-7DOI
  46. [51]Perera, B. B. P., Stappers, B. W., Babak, S., et al. 2018, MNRAS, 478, 218, doi: 10.1093/mnras/sty1116DOI
  47. [52]Peters, P. C. 1964, Physical Review, 136, 1224, doi: 10.1103/PhysRev.136.B1224DOI
  48. [53]Petreczky, P. 2012, Journal of Physics G Nuclear Physics, 39, 093002, doi: 10.1088/0954-3899/39/9/093002DOI
  49. [54]Planck Collaboration, Aghanim, N., Akrami, Y., et al. 2020, A&A, 641, A6, doi: 10.1051/0004-6361/201833910
  50. [55]Punturo, M., Abernathy, M., Acernese, F., et al. 2010, Classical and Quantum Gravity, 27, 194002, doi: 10.1088/0264-9381/27/19/194002DOI
  51. [57]Reitze, D., Adhikari, R. X., Ballmer, S., et al. 2019, in Bulletin of the American Astronomical Society, Vol. 51, 35, doi: 10.48550/arXiv.1907.08338
  52. [58]Renzini, A. I., Goncharov, B., Jenkins, A. C., & Meyers, P. M. 2022, Galaxies, 10, 34, doi: 10.3390/galaxies10010034DOI
  53. [60]Sazhin, M. V. 1978, Soviet Ast., 22, 36
  54. [61]Sesana, A., & Figueroa, A. 2025, arXiv e-prints, arXiv:2512.18822, doi: 10.48550/arXiv.2512.18822
  55. [62]Sesana, A., Korschorkova, N., Sedda, M. A., et al. 2021, Experimental Astronomy, 51, 1333–1383, doi: 10.1007/s10686-021-09709-9DOI
  56. [63]Shannon, R. M., Bhat, N. D. R., Challameau, A., et al. 2025, The Open Journal of Astrophysics, 8, 54243, doi: 10.33232/001c.154243DOI
  57. [64]Shapiro, S. L., & Teukolsky, S. A. 1983, Black holes, white dwarfs and neutron stars. (The physics of compact objects (John Wiley & Sons, Ltd)), doi: 10.1002/9783527617661DOI
  58. [65]Sheridan, C., & Keane, E. 2026, Investigating the Viability of Dedicated Microhertz Gravitational Wave Pulsar Timing [dataset], Zenodo, doi: 10.5281/zenodo.2369055arxiv.org/abs/2611.12179
  59. [66]Thrane, E., & Romano, J. D. 2013, Phys. Rev. D, 88, 124032, doi: 10.1103/PhysRevD.88.124032DOI
  60. [67]Wang, S. Q., Wang, N., Wang, J. B., et al. 2024, ApJ, 964, 6, doi: 10.3847/1538-4357/ad327b
  61. [68]Witten, E. 1984, Phys. Rev. D, 30, 272, doi: 10.1103/PhysRevD.30.272DOI
  62. [71]Zel’dovich, Y. B., Kobzarev, Y. I., & Okun’, L. B. 1974, Zhurnal Eksperimentalnoi i Teoreticheskoi Fiziki, 67, 3
  63. [72]Zic, A., Reardon, D. J., Kapur, A., et al. 2023, PASA, 40, e049, doi: 10.1017/pasa.2023.36DOI
  64. [73]Zonca, A., Singer, L., Lenn, D., et al. 2019, Journal of Open Source Software, 4, 1298, doi: 10.21105/joss.01298DOI

Paper details

vlnc:W_byvwh4tj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents