Arxiv:2007.11554V1 [Astro-Ph.IM] 22 Jul 2020

Relaying Swarms of Low-Mass Interstellar Probes$

David G Messerschmitta, Philip Lubinb, Ian Morrisonc

aUniversity of Calfornia at Berkeley, Department of Electrical Engineering and Computer Sciences, USA bUniversity of California at Santa Barbara, Department of Physics, USA cCurtin University, International Centre for Radio Astronomy Research, Australia

Abstract Low-mass probes propelled by directed energy from earth are an early option for exploration of nearby star systems. A challenging aspect of such technology is returning scientific observational data to earth. We compare two configurations for achieving this. A direct configuration utilizes optical transmission from the probe to a terrestrial receiver employing a large photon collector. In a relay configuration, probes spaced at uniform intervals act as regenerative repeaters for the scientific data, which eventually arrives at a terrestrial receiver from the most recently launched probe. A number of advantages and disadvantages of the relay configuration are discussed. A numerical comparison approximates equal probe mass in the two cases by using the same optical transmit power and equivalent total transmit plus receive aperture area. When the total downlink data rate is equal, the relay configuration benefits from a smaller terrestrial receive collector, but also requires very frequent launches to achieve higher data rates due to the limitations on relay probe receive aperture area. The direct configuration can achieve higher data rates without such frequent launches by increasing terrestrial collector area. A single-point failure problem in the relay configuration can be addressed by introducing relay-bypass modes, but only at the expense of further increases in launch rate or reductions in data volume, as well as a considerable increase in design and operational complexity. Taking into account launch and collector area costs, the direct configuration is found to achieve lower overall cost by a wide margin over a range of cost parameter values and data rates. Keywords: Interstellar communications low-mass space probes arXiv:2007.11554v1 [astro-ph.IM] 22 Jul 2020

$Copyright c 2020

Preprint submitted to Elsevier July 23, 2020 Nomenclature other hand, multiple generations of a digital representation of information are almost totally free

λ0 Wavelength of optical communication of these accumulating impairments. There may be u0 Speed of all probes some impairment due to occasional errors in the T∆ Time interval between probe launches recovery of a digital representation, but these can u0 Speed of all probes usually be rendered insignificant with little penalty. P0 Probe transmit optical power A0 Effective area of a transmit aperture in direct con- In communications, the idea is to periodically figuration position regenerative repeaters, which partitions the V0 Total scientific data downloaded from each probe end-to-end communication into links. Repeaters D0 Probe to earth distance at start of downlink operation in direct configuation encompass receiver/transmitter combinations that D1 Probe to earth distance at completion of of down- recover the original digital data almost error-free link operation in the direct configuration and retransmit it free of any noise and distortion Dl Probe to earth distance at end of launch (start of introduced on the previous link. This simple idea relay operation) BPP Photon efficiency of communication (bits reliably largely prevents the accumulation of noise and dis- recovered per detected photon) tortion effects over long distances. Periodic regen- Po Probability of outage due to atmospheric impair- eration is the primary reason that local vs long- ments (weather and sunlight scattering) distance voice and video calls are today largely in- Jr Maximum number of missing or failed probes which can be bypassed in the relay configuration distinguishable in quality. Regeneration is the original (and probably still the single most important) motivation for migration from analog to digital trans- 1. Introduction mission [4], as well as for the long-term storage of 1 Previous studies of communication of scientific various media in digital rather than analog form. data from low-mass interstellar probes have assumed Just as regenerative repeaters are applicable to direct probe-to-earth transmission [1, 2, 3]. An terrestrial communications, they have been used in alternative that invariably arises in discussions of space communication as well, where the term relay the data downlink system configuration is the idea is applied to a single regenerative repeater node. A of opportunistically tasking more recently launched common configuration uses an orbital satellite as probes as relays servicing the downlink communica- a relay for a surface vehicle or other orbital satel- tion needs of probes launched earlier. Particularly lites in the vicinity of Earth [5], the Moon [6], and attractive is the opportunity to devote otherwise- Mars [7]. This has the benefit (relative to direct-to- unused electrical power resources and communica- earth transmission) of reducing the surface vehicle’s tion capabilities to the relaying function while probes resources (aperture area, electrical power, etc) de- traveling to the target star, during which time the voted to communication. opportunities for scientific observations are more 1.2. Application to low-mass probes limited. We quantitatively explore the merits of the re- 1.1. History lay idea as applied to low-mass interstellar probes by considering and comparing two alternative con- The basic idea behind this relaying of data has a figurations for a swarm of probes performing a flyby long and storied history in the digital transmission of a target star. These configurations are: and storage of data. The periodic regeneration of digitally-represented data is a powerful technique Direct configuration. The probes operate com- used widely in terrestrial communication systems pletely independently of one another. Each as well as in storage systems. In such systems the probe serves as a scientific data collection plat- data is inevitably copied to create new replicas or form during encounter with the target star, to replace an aging copy. Each such copy is called a as well as incorporates a dedicated and sepa- new generation, which is the origin of the term re- rate communication downlink for that scien- generation. With multiple generations of a medium tific data which transmits directly like audio or video that is communicated or stored in analog (continuous in time or amplitude) form, 1 In fiber optics systems, due to technological opportu- successive generations inevitably deteriorate due to nities like low-loss and low-dispersion fibers, regenerative the accumulation of noise and distortion. On the repeaters are often replaced by simpler optical amplifiers, which harks back to the practice in earlier analog systems. 2 post-encounter to a terrestrial receiver. Such a receiver will typically receive data from mul- (a) Direct tiple post-encounter probes simultaneously, and thus multiple downlinks operate concurrently. Earth Target star Post- This configuration was analyzed in [1]. encounter (b) Relay Regeneration Probes Relay configuration. The probes operate independently in their scientific observations, but work cooperatively to return the resulting sci- Earth Near+relay Encounter+ relay probe entific data to earth. There is a single down- probe Relay probes link originating from a post-encounter probe with scientific data to share with earth. The Figure 1: Comparison of the direct and relay configurations. downlink then passes through regenerative re- (a) Multiple probes communicating directly with earth may peaters carried by all more recently launched share a single receive collector, which separates the probe probes. In each repeater, the scientific data signals in time or frequency or spatially. (b) Probes are assumed to be launched at regular intervals, and moving embodied in the downlink is recovered at the at uniform velocity u0 and thus have uniformly spaced dis- highest available fidelity before being re-encoded tance from earth. Each post-encounter probe transmits its and re-transmitted to another probe closer to scientific data on the downlink, one at a time. That single downlink passes through each and every probe that is closer earth. to earth. Each probe in the relay operates as a regenerative repeater, recovering a faithful version of the scientific data These configurations are illustrated in Fig.1. In the and then retransmitting to another probe closer to earth or direct configuration, multiple post-encounter probes (in the case of a near probe) directly to a receiver on earth. are involved in downlink communication, but they have no other function during downlink operation. In particular, their scientific observations have been figuration but benefiting from a much shorter completed during a previous encounter with the propagation distance. As a result of that shorter target star, so that each probe carries both scien- distance, the area of the terrestrial receive col- tific instrumentation and communication capabil- lector can be considerably smaller. ity. Since the probes operate independently they • Following the launch of another probe but still can be heterogeneous (in launch interval, mass/velocity, pre-encounter, the relay probe has a single data rate, scientific mission etc). function, which is to serve as a regenerative In the relay configuration the probes are launched repeater in support of the downlink. at fixed intervals, and are homogeneous in mass/velocity so that they cruise with unchanging inter-probe dis- • During its subsequent encounter with the tar- tance. Their scientific instrumentation and mis- get star, the relay probe performs its scien- sions may still be differentiated, but there is no op- tific observations and concurrently performs portunity to adjust probe mass to instrumentation the relay function (with its regenerative re- and data volume needs as proposed in [1]. peater in operation). Also in the relay configuration each probe de- ploys a communication receiver (as well as transmitter) and performs the regeneration function. From 1.3. Comparison a communication perspective, the probes perform The goal here is to understand the performance four distinct functions during their lifetime. There limitations and cost implications of the two config- is a distinction between a relay probe and a post- urations, and to compare them. This comparison encounter probe, as the relay probe is required to is divided into two categories: technical and eco- receive downlink data from a more distant probe, nomic. To approximate a fixed probe mass in all regenerate that data, and then retransmit it to a cases, the transmit power and total aperture areas probe closer to earth. There are three distinct op- are held fixed. From there the technical and eco- erational phases for relay probes: nomic comparisons diverge: • The most recently launched relay probe (called • In the technical comparison the goal is to ap- the near relay probe) transmits directly to a preciate the difference in terrestrial aperture terrestrial receiver, similarly to the direct con- 3 area due to the much greater propagation dis- the design and operation. In the direct configura- tance in the direct configuration. To accom- tion each probe operates independently, and launch plish this, the launch interval and total scien- failures affect a single probe. However, the system tific data volume conveyed to earth by each remains vulnerable to generalized shutdowns when probe are forced to be the same, and the ter- there are failures or problems with the terrestrial restrial collector areas are compared. This receiver, which is shared across all probes currently does not account for any difference in total in their post-encounter phase. data latency. Ultimately cost considerations will be important in choice of a configuration (see §6 and §7). There • In the economic comparison, it is important is a tradeoff between larger launch costs (relay con- to allow the launch frequency and interval to figuration) and the capital cost of a larger receive be different for the direct and relay configu- terrestrial collector (direct configuration). For the rations. This is because the direct configura- relay configuration, achieving higher data volumes tion derives economic benefit from reducing cumulative across all probes requires an investment the number of probes launched and compen- in a shorter launch interval, with the attendant re- sating for this with a larger data volume per current energy costs of launches. The data volume probe, and this is not an option in the relay per probe is severely limited, which may be in con- configuration. For a fair comparison, the to- flict with scientific requirements. Launches must be tal cumulative downlink data rate is held fixed frequenct, regular, and uninterrupted for the entire between the two configurations, and the costs life of the system, which precludes commensal uses are compared. Only costs that differ strongly of the launch infrastructure. On the other hand, between the two configurations, namely the replication of the receiver for parallel uses is less cost of launch and the cost of collector area, expensive due to the small collector aperture size. are considered. Achieving higher data volumes for the direct configuration requires a one-time capital investment 1.4. Summary of outcomes in a significantly larger terrestrial receive collector, The advantage of the relay configuration is a sig- which is compensated by less frequent launches. nificantly smaller receive terrestrial collector area, The launch interval and probe mass can be freely not much larger than the receive aperture on a relay varied according to scientific objectives or advanc- probe (see §4) The price paid for this advantage is ing technology. Commensal uses of the launch in- a small data volume per probe, constrained by the frastructure and cost sharing across different mis- receive aperture size on each relay probe. The prin- sions objectives (such as outer planet and a mul- ciple technical advantage of the direct configuration titude of nearby stars) is feasible due to a flexible is the possibility of much larger data volumes per launch schedule. An estimate of total costs and probe, because the receive terrestrial collector area unit costs finds that overall the direct configura- is not constrained by probe mass considerations and tion is more cost-effective by a wide margin across can therefore be much larger. In the direct config- a range of cost parameter values and downlink data uration, the data volume per probe can be freely rates. matched to science requirements by adjusting the Overall, taking into account both technical and terrestrial collector area, but in the relay configu- cost considerations, the direct configuration appears ration the data volume is constrained by the probe to have overwhelming advantages. parameters (and hence probe mass considerations). The relay configuration is much more complex from a design and operational perspective. In addi- 2. Technical comparison tion it is quite vulnerable to launch and probe fail- From a technical perspective, the relay configu- ures, with individual missing or failed probes com- ration has advantages and disadvantages. promising the operation of the entire system (see §5). To an extent this can be overcome by building 2.1. Advantages in robustness to failure, but only at a significant cost in terms of reduced data volumes per probe as Near terrestrial receiver. Since a terrestrial re- well as some significant additional complications in ceiver on earth receives downlink data from the most recently launched probe, the short

4 propagation distance from that probe implies all the scientific data it has accumulated over that this receiver’s collector area is signifi- a limited time (equal to the inter-launch inter- cantly smaller. In fact the terrestrial receiver val), and subsequently shuts itself down per- is comparable in its performance metrics to manently. Thus, the lifetime of the electri- a relay probe receiver, with the added com- cal power source is advantageously fixed and plication that it has to deal with atmospheric short. impairments and with the additional propagation distance that accounts for the launch. Fixed power and distance. The propagation of all signals between relay and post-encounter Outage mitigation. The relay probe transmit func- probes is the same. Only the near-link probe tion does not need to account for atmospheric need compensate for variations in propaga- outages (sunlight and weather). Only the near tion distance, and that is as simple as design- probe has to communicate through the at- ing for the worst-case. However, in practice mosphere, and thus suffer the additional re- improving the system reliability by allowing dundancy overhead to counter outages due to for relay probe bypass (see §5) negates this weather, sunlight scattering, etc. This larger advantage. overhead can be compensated by a modest increase in the area of the terrestrial collector. 2.2. Disadvantages Multiple functionality. Although the communi- Wavelength flexibility. With the unique excep- tion of the near relay probe, the transmit wave- cations functionality is distinctive between post- length can be chosen freely without regard to encounter and relay probes, each and every atmospheric transparency. This could result probe serves all roles (scientific and communications) at different times during the mission in smaller transmit/receive apertures, and op- and thus has to embody all the requisite ca- eration at UV wavelengths would in addition pabilities with all the implications of that to significantly reduce the background radiation 2 originating from the target star.2 mass, electrical power, attitude control, etc. Simple multiplexing. The complexities of multi- Multi-probe data. For the direct configuration, there are parallel downlinks operating concur- plexing signals from multiple probes and sepa- rently. In the relay configuration, there is a rating them at the receiver are avoided. While single downlink that is used sequentially by the single downlink is time-shared between all the probes, in the simplest case assumed here each probe post-encounter to convey its sci- only one probe at a time makes use of the entific data back to earth. All else equal, the downlink. Even if more than one probe orig- data rate on that single downlink is significantly higher, with all the implications to pro- inates downlink data concurrently, the time- cessing and electrical power requirements.2 division multiplexing this implies is easy to implement. Concurrent observation and regeneration. In Limited coverage. Each relay probe and the re- the direct configuration the scientific observa- lay near-link terrestrial receiver require lim- tions and downlink communications need not be concurrent, and this moderates the electri- ited coverage because they transmit to a sin- cal power requirements. In the relay configu- gle probe or terrestrial receiver, or receive from a single probe. This reduced coverage in- ration a probe performing scientific observa- creases aperture sensitivity, and may also sim- tions must concurrently serve as a regenera- plify tracking. tive repeater for the downlink serving a post- encounter probe. Thus the electrical power Limited transmission time. In the relay config- requirements are correspondingly higher, al- uration, each post-encounter probe transmits though this may possibly be offset by photo- voltaic power generation available in the vicinity of the target star.2 2 The probe mass implications, coming from example a aperture size, electrical power generation, and processing re- Smaller data volume We will find that the di- quirements, are not considered in our comparisons. rect configuration can achieve considerably 5 greater data volumes per probe. The reason Continuous launches. Probes must be launched for this is that the receive terrestrial collector at regular intervals, without breaks for main- is unconstrained in size, whereas in the re- tenance, weather events, etc. Any interrup- lay configuration the receive aperture on each tion in the launch schedule cuts off the scien- relay probe is highly constrained by probe tific data downlinks from all previously launched mass limitations. Within one relay-to-relay probes unless there is a probe bypass capabil- link, the small apertures for both transmit ity (see §5) which substantially reduces data and receive functions are not overcome by the volumes. shorter propagation distance unless the inter- Inter-probe interference. Earlier-launched probes launch interval is very short, in which case the transmitting to their respective relays will be launch energy costs become very large. a source of unwanted interference with relay Coupling of launch interval and data volumes. links closer to earth. With considerable com- In the direct configuration, the number of probes plication this interference would be largely and data volume per probe can be chosen in- eliminated by using different wavelengths on dependently based on maximization of scien- the different links.2 tific return. In the relay configuration these Regeneration processing. Although all-optical re- two parameters are closely coupled and de- generation of signals has been studied for fiber pendent due to the relationship between data systems [8], in a photon-starvation mode with volume and launch interval. high photon efficiency neither optical ampli- Speed uniformity. All probes must travel at the fication nor optical regeneration is an option. same velocity to maintain a consistent maxi- Thus, each relay probe is a full regenerative mum propagation distance. Thus any oppor- repeater which performs optical-to-electrical tunity to trade larger probe mass for lower conversion and implements the complete re- speed, greater data latency, and higher data ceive and transmit stacks (modulation decod- volume as in the direct configuration [1], is ing, error-correction decoding, error-correction abandoned. coding, modulation coding). The processing required for error-correction decoding in par- Parallax and pointing complications. Since ticular is quite significant. This processing is probe launches occur at different times in the replicated in every probe, and may consume earth’s orbital cycle, and are aimed in slightly an electrical power that is significant in com- different directions to track the proper motion parison to the optical transmit power.2 of the target star, the trajectories of probes are not co-linear. The pointing of each relay Error accumulation. In the relay configuration probe to a probe nearer to earth (rather than bit errors in recovery of scientific data will ac- earth itself) will be complicated by parallax cumulate through multiple regeneration steps. effects, and further there is no apparent refer- Thus, the permissible error rate objective in ence for this pointing function. In the direct each relay link will have to be correspondingly configuration, the sun provides a convenient tighter, resulting in slightly higher transmit reference for pointing back to earth. powers. Simultaneous attitude adjustment. In the di- Timing recovery. High photon efficiency using rect configuration, attitude adjustment for sci- pulse-position modulation (PPM) with pho- entific observations and for data communica- ton starvation presents a difficult challenge in tions can be separated, since the probe need deriving the timing of the PPM slots. This not do both functions simultaneously. For the would presumably have to be performed in relay configuation, a probe performing atti- real-time (as opposed to post-processing fol- tude adjustments for its own scientific obser- lowing completion of the mission as in the vations has to simultaneously maintain point- direct case). It is likely this would only be ing accuracy in both its data transmission and feasible at higher received power levels and reception functions. This presumably implies poorer photon efficiencies, reducing data rate. independence of aperture pointing and probe It would also consume considerable processing attitude.2 power.2 6 Cryogenic. Cryogenic (low) temperatures are de- Table 1: Comparison of design parameters and metrics sired for the receive photonics and optics. This Parameter Direct Relay will likely be accomplished with passive cool- Not-near Near ing in general as the probes will run cold dur- Propagation D0 to D1 T∆ · u0 Dl + T∆ · u0 distance ing the cruise phase, but may be a challenge Average trans- P0 P0 P0 T in the vicinity of the target star. The direct mit power PA Transmit aper- Probe Probe Probe configuration requires no receiver and thus no ture location cryogenic temperatures. Transmit aper- A0 A0/2 A0/2 ture effective T area Ae Numerous single points of failure. The failure Receive collec- Terrestrial Probe Terrestrial of a single probe results in partial or complete tor location Receive aper- Ad A0/2 An mission failure for all earlier-launched probes. ture or collector This can be offset by adding the complication total area of bypassing probes, although this comes at the expense of lower data volumes. Equal probe resources. In the interest of a com- Complexity. A relay approach will impose con- parison emphasizing the distinctions between siderable design and operational complexity. the two configurations, parameters of the probe In contrast to planetary missions there is no are fixed to approximate equivalent mass probe opportunity to upgrade software, and thus ex- requirements. In particular, the transmit power tensive operational testing will be required P0 and the total probe aperture area (for trans- before the first launch. Aspects of the opera- mission and reception) A0 are kept equal. tion such as attitude adjustment would probably be impossible to test in advance, raising Terrestrial collector area. The receive terrestrial the risk of malfunction. collector in the near relay case has a larger area An > A0/2 than the other relay probes Many abandoned probes. At the end of the sys- due to a propagation distance that is larger tem life, a full cohort of relay probes will be by launch distance Dl. Near relay probe re- precluded from communicating data back to lay downlink operation is assumed to begin earth for lack of subsequent relay capability. following the completion of launch, and hence the propagation distance is larger by Dl. 3. Basis of comparisons Downlink data rate. To approximate a comparable scientific return for the two configura- The quantitative results in §4 and §6 focus on tions, the total cumulative downlink data rate a comparison between the terrestrial collector size across all probes is assumed to be the same and system cost for the direct and relay configu- in the direct and relay configurations. rations. For both configurations, the quantitative models for data rate and data volume are adopted from [1]. The assumed parameters are listed in 3.1. Comparison shortcomings Tbl.1 for three cases: Direct, and in the case of a Our comparison of the two configurations ne- relay configuration, a relay-to-relay and near-relay- glects several factors that will differ: to-earth downlink. The chosen parameters are notable in several Electrical power. The electrical power is consumed respects: in functions other than communications, including attitude control and data processing. Propagation distance. The launch interval T∆ is In particular, in the relay configuration each a major consideration in the relay configura- probe has to have adequate electrical power tion, since it governs the inter-relay distance for concurrent scientific observation and com- and hence the downlink data rate. The propa- munications, including attitude control, but gation distance is greater than the target star not in the direct configuration. The data rates distance D0 for the direct configuration, and are also considerably higher in the relay con- is generally far smaller in the relay configura- figuration, and this will consume additional tion. processing electrical power. 7 Processing overhead. We expect the processing Table 2: Numerical parameters chosen for comparison overhead to be considerably higher for the re- Parameter Value lay probes because of their higher data rate Distance D0 4.24 ly (all else equal) and because of the significant Direct rate of propagation distances D1 − D0 0.424 ly processing associated with regeneration (prin- Wavelength λ0 400 nm Average transmit power P0 100. mW cipally receiver error-correction decoding). Photon efficiency BPP 10.9 bits/ph 2 Aperture area A0 100 cm Error rate objectives. Due to the accumulation Atmospheric outage probability 0.58 of errors through multiple regenerative repeater stages, the error rate objective has to be tight- which conduct scientific observations and re- ened to account for an increase in regenera- turn the resulting scientific data. The number tion stages. Since the models used here are of probes and the data volume per probe V0 based on theoretical limits on communication determine the total data rate b. For a given with reliable recovery of scientific data, this b, as T∆ is adjusted the granularity of the issue is not directly relevant. However, in partitioning of that data between probes (as practice tighter error rate objectives will in- measured by V ) changes. crease the transmission power requirement 0 slightly. • For the relay configuration, b and V0 are determined by the probe parameters and by T∆, One-at-a-time downlink capture. It would be which determines the inter-relay propagation feasible for more than one post-encounter probes distance. For the direct configuration, for any to access and share the single relay down- assumed T∆ the collector area Ad has to be link concurrently. However, as pictured in adjusted to match b and T . Fig.1 it is assumed that a single probe at a ∆ time accesses the downlink. For a fixed cu- For a fixed b, the primary difference between the mulative downlink data rate, this simplifica- two configurations is the T∆ as well as the area of tion does not affect the total data volume V0 the terrestrial receive collector. downloaded from each probe, but it does ad- For purposes of a numerical example, we adopt versely impact a probe’s ability to do inter- the specific values listed in Tbl.2 for the probe pa- leaving to counter atmospheric outages on the rameters defined in Tbl.1. As a basis of comparison near-link.3 there are at least two approaches. In the technical comparison of this section not only the data rate b These factors all serve to make the relay configura- but also the launch interval T∆ are constrained to tion less attractive than suggested by the numerical be the same for the two configurations. The moti- results to follow. vation here is to isolate the effect on the collector area metric by fixing all other parameters and mak- 4. Technical comparison ing them identical. Naturally we will find that this collector area is larger for the direct configuration Our basis for a quantitative comparison between because its signal is propagating a greater distance. the direct and relay configuration is to set the total This makes it appear that the relay configuration cumulative data rate b (with units of bits/second) is advantageous. Subsequently in the cost compar- equal for the two configurations, as well as the probe ison, the constraint that T∆ is the same is removed parameters. There are two primary parameters that (see §6). In that case, we find that the direct con- will differ: figuration becomes strongly advantageous from an economic (as opposed to technical) perspective. • The launch interval T can be varied. This ∆ The collector area for the direct and relay con- controls the number of probes per unit time figurations are plotted in Fig.2, and compared to the area of the receive aperture on a relay probe. 3 Countering atmospheric outages could still be per- The collector area for the direct configuration is formed in later regenerative repeaters, but this would require considerably larger, and that difference grows with one probe (and hence every probe) to have sufficient storage b. As b increases a shorter launch interval T∆ is for buffering a substantial quantity of data. required, and the resulting number of probes in the

8 9 ) 2 4 8

2 direct & relay 7 Ad for near direct collector

0 An for near relay collector 6 aavlm nbits ) in volume ( data

-2 10 5 log olco rai m in area aperture / collector = ( A 10 -4 log A0/2 for relay probe collector 4

3 -2 0 2 4 6 -2 0 2 4 6

log10(b= cumulative data rate in bits/sec } log10(b= cumulative data rate in bits/sec }

Figure 2: A log-log plot of the received aperture or collec- Figure 4: For the same conditions as Fig.2, a log-log plot of tor area vs b, which is the total downlink data rate cumula- the data volume V0 for each probe. This is the same for the tive over all probes concurrently operating their downlinks. direct and relay configurations, since the launch interval T∆ The range of data rates is nine orders of magnitude, from is constrained to be the same in both cases. The increase b=1 mb/s to b=1 Mb/s. The direct and near relay terrestrial in V0 with b is achieved by decreasing the propagation dis- collector areas are included, as well as the aperture area on tance between relay probes in the relay configuration. In the a relay probe. The relay probe parameters are constrained direct configuration, larger V0 is achieved by increasing the across all b, and thus the relay probe aperture area is fixed. terrestrial collector area as shown in Fig.2. The near relay collector area is slightly larger than the relay probe aperture in order to compensate for atmospheric outages. The direct configuration collector area is considerably Also as b increases the data volume per probe larger due to the greater propagation distance compared to the near relay probe. V0 increases as shown in Fig.4. Thus, the larger b is accommodated by simultaneously increasing the number of probes and the data volume V for each 7 0 probe. This coupling in V0 and b is a disadvantage 6 of the relay mode, due to the inflexible partitioning

direct & relay of observations and resulting data volume among 5 probes.

4 This comparison in which the launch interval T∆ is constrained to be equal for the two conﬁg- 3 urations favors the relay conﬁguration due to its none probes ) pre - encounter of ( number

10 smaller terrestrial collector area. However, the di- 2 log rect conﬁguration oﬀers the freedom to increase T∆

1 and compensate for the fewer probes with a larger -2 0 2 4 6 data volume V0 per probe (achieved by increasing log (b= cumulative data rate in bits/sec } 10 the terrestrial collector area). Due to the substan- Figure 3: For the same conditions as Fig.2, a log-log plot tial cost of each launch, this is strongly advanta- of the number of probes in their pre-encounter trajectories. geous from a cost standpoint (see §6). This is the same for the direct and relay conﬁgurations, since the launch interval T∆ is constrained to be the same in both cases. As the cumulative data rate b increases, a smaller T∆ 5. System reliability is required, with a resulting increase in the number of probes. In order to achieve b=1 Mb/s, a launch interval T∆=5.1 min In the straightforward relay system pictured in is required, which would likely necessitate more than one directed-energy beamer. For the relay conﬁguration it is Fig.1, even a single missed launch or a single failed assumed that Jp=1. probe will render useless and forever abandoned all probes further from earth than the point of missing or failed probe. This is clearly an unacceptable level pre-encounter phase of their trajectories increases of system reliability. Consider for example a relay as shown in Fig.3. At higher data rates the launch system in which there are Jt relay probes at any rate becomes impractical for a single launcher, and given time. Whenever one or more of these probes the energy cost of those launches will be large. fails or is not successfully launched, then the down-

9 A mandatory adjustment to data rate for the ? missing or failed link requires knowledge of the failure at the transmitter so that it can adjust its transmit energy per PPM pulse (likely by increasing pulse duration) to compensate for the greater propaga- Figure 5: When a single probe is missing or has failed, this tion distance and still achieve the expected number can be overcome by bypassing that probe, transmitting in- of photon detections per PPM slot at the receiver. stead to the probe following immediately behind the missing or failed probe. A probe has no means of knowing about failures or missing probes closer to earth based on its own observations, although it can detect failed or miss- link data from all probes that have completed scien- ing probes farther from earth due to the absence of tific observations and begun downlink transmission received signal. are abandoned, as are the potential observations As a result probe bypassing adds an additional from all relay probes previously launched. requirement that probes closer to earth inform probes farther from earth of intermediate missing or failed 5.1. Single point of failure probes. This required level of coordination between We define as a system failure any situation where probes could be obtained by a telemetry transmis- multiple probes must be abandoned. When a single sion from the operational probe closer to earth, missing or failed probe can cause a system failure, which can observe the failure directly, requesting the probability of this event is that transmission be redirected to it. Thus, each probe requires a limited transmission and receiv- Jt Pr {system failure} = 1 − (1 − p) ≈ Jt · p (1) ing capability for control telemetry communication in the direction away from earth, adding two ad- where p is the probability of a single probe fail- ditional apertures and associated transmitter cir- ure and probe failures are assumed to be statisti- cuitry and electrical power consumption to the probe cally independent. The approximation is valid for mass budget. In the following we neglect the ad- Jt · p 1. Thus p has to be kept small, and de- verse mass implications of this enhanced capability. creasing the launch interval in the interest of greater More generally Jr ≥ 1 failed or missing probes data volume V0 either imposes a smaller p or alter- can be bypassed, increasing the transmission dis- natively is deleterious to system reliability. tance by a factor of Jr+1. The same cumulative data rate can be achieved if the launch interval T∆ 5.2. Probe bypassing to improve reliability is reduced by the factor Jr+1, thereby restoring A way to improve system robustness and im- the original propagation distance. This increases prove reliability is to configure the system so that the frequency of launches, which given the signifi- it can tolerate Jr ≥ 1 individual probe or launch cant expense of launches is a substantial penalty in failures with degraded performance metrics (as op- system cost. An alternative is to retain the original posed to outright system failure). One way to ac- launch frequency but reduce the data rate of the 2 complish this is illustrated for Jr=1 in Fig.5. In downlink by a factor of (Jr+1) to account for the this configuration, a single missing or failed probe is greater transmission distance. bypassed by transmitting to the next downstream probe. This requires a mandatory downward ad- 5.3. Probability of system failure justment in the data rate for this individual link The expectation of improved system reliability due to the greater propagation distance. This link with an increase in the number of bypassed probes then becomes a bottleneck, reducing the data rate Jr is confirmed by a calculation of the probability until the missing and failed probe is eliminated from of that event. the relay system because it has (or would have) Suppose we have a system configuration in which moved into the post-encounter phase. Fortunately Jr contiguous probe failures can be accommodated multiple single-probe failures can be accommodated with the bypass of Jr probes in the relay system. without any additional deterioration in data rate, The system as a whole then fails (meaning that but only so long as these failures are isolated from the supported data volume V0 per probe goes to one another, and not contiguous. zero for multiple probes) if Jr+1 or more contiguous probes fail. In that case, the propagation dis- 10 0 The primary conclusion of Fig.6 and Fig.7 is that the probability of system failure can be ren- -5 1 dered much smaller than the probability of probe 2 -10 failure. In other words, a reliable system can be 3 constructed of unreliable probes by building in tol- -15 4 erance for missing or failed probes. As the reli- 5 ability of individual probes and their launches is -20 6 rbblt fsse failure ) system of ( probability

10 7 improved, the value of Jr can be reduced. -25 log 8 -30 6. Economic considerations -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 log10(p = probability of probe failure) The preceding analysis has focused on the technical considerations in a comparison between the di- Figure 6: A log-log plot of the probability of system failure vs the probability of probe failure p for different values of rect and relay configurations. Following from these 1≤Jr≤8, with the curves labeled by Jr. A system failure is technical considerations are major implications to defined as more than Jr contiguous missing or failed probes. the economic structure of the low-mass probe project. This assumes there are Jt=200 probes in total. There are three principle assets in the system: The ground infrastructure for launch and data recep- 0 tion, the relay probes, and the scientific return in terms of scientific observational data returned reli- -2 ably to earth.

-4 10 000 6.1. Fixed and recurring costs and returns 100 -6 1000 The relative costs and returns are strongly af-

rbblt fsse failure ) system of ( probability fected by the choice of direct or relay conﬁguration:

10 -8 log Directed-energy launch. The up-front fixed cost -10 of the terrestrial launcher infrastructure is dom- -3.0 -2.5 -2.0 -1.5 -1.0 inated by the peak energy delivered to a probe’s log10(p = probability of probe failure) sail, and related to that the velocity of the probe and data latency [3]. This fixed cost is Figure 7: A repeat of Fig.6 except that Jr=3 is kept fixed 2 3 4 and three values of Jt={10 , 10 , 10 } are plotted. Each or- not expected to be materially affected by the der of magnitude increase in the number of probes increases choice of the direct vs relay configuration. the probability of system failure by approximately an order of magnitude. Pre-encounter probes. The relay configuration will generally require significantly more frequent launches if a similar scientific data vol- tance to the next functional probe is greater than ume per probe is to be maintained. The en- the design limit. Suppose too that the total number ergy cost of each launch is significant, result- of probes in a relay system is J . If J J then t t r ing in a significantly larger recurring cost over- multiple failures of up to J contiguous probes are r all. The recurring cost per unit of data vol- allowed without precipitating a system failure. ume is higher, and strongly dependent on the A formula for system failure probability is cited launch interval T . in §Appendix B assuming that each probe fails in- ∆ dependently with probability p. The probability of Receive terrestrial collector. For the direct con- system failure is plotted in Fig.6 and Fig.7 as a func- figuration the fixed cost of the collector will tion of the probability p of probe failure. In Fig.6 be roughly proportional to its area, and hence the value of Jr, the maximum number of probes to the data volume returned from each probe. that can be bypassed, is varied while the number of For the relay configuration the terrestrial col- probes is fixed. In Fig.7 the value of Jt, the total lector area is essentially invariant to data vol- number of probes in the system, is varied while Jr ume since it is constrained by the probe mass. is kept fixed.

11 Science return. The direct configuration can gen- configuration is adopted. Thus generally we erally return a larger data volume from each expect that a separate receive infrastructure, probe, and hence the recurring cost per unit with its substantial fixed costs, will be ded- of scientific data will be lower. On the other icated to each separate target. In this case hand the relay configuration probe swarm will the relay configuration has an advantage in generally fly a larger number of probes by the its considerably smaller collector area, which target, increasing the total scientific return makes it more economically attractive to in- in proportion to the number of probes (as- vest in multiple receivers for multiple targets. suming that the scientific data can be more fragmented across probes without reducing its In summary, the direct configuration is more fa- value). vorable in its support for commensal missions with respect to the launcher, while the relay configura- In summary, the direct configuration will generally tion is more favorable with respect to the terrestrial have a significantly larger fixed cost and signifi- receiver. cantly lower recurring cost than the relay configuration. These costs are compared numerically in 6.4. Long-term reuse §7. A second even longer-term issue is the non-commensal reuse of investments made in the launcher and ter- 6.2. Risk restrial receiver for new science missions in the fu- The relay configuration definitely increases the ture. This opportunity favors a solution with higher risk involved in infrastructure investments due to fixed costs (since these costs are amortized over the greater complexity and greater number of fail- more usage) and lower recurring costs (so that each ure modalities (see §5). The performance metrics usage is less expensive). Thus it favors the direct are generally rendered less favorable by technical configuration. measures taken to reduce or limit these risks. 7. Numerical cost comparisons 6.3. Commensal usage In addition to fixed and recurring costs related A major difference between the direct and re- to exploration of a single target, there are longer- lay configurations is the cost structure. To achieve term issues related to the commensal utilization of comparable scientific return (measured here by data the fixed-investment ground infrastructure. These rate), the relay configuration requires more frequent differ considerably for the two configurations. launches with less data volume returned per probe, while the direct configuration is able to reduce the Launcher. The directed-energy launcher in the re- frequency of launches but increases the area of the lay configuration will be generally preoccu- receive collector. The only criterion which can es- pied with keeping a regular repetitive launch tablish the relative merits of these two approaches cycle for probes to a single target. Thus, the is the relative cost. In particular for the direct con- opportunity for commensal usage to serve a figuration this can answer the quandry “as a way to second target will be severely limited. The di- increase the scientific data return, which option is rect configuration offers more opportunity for more cost effective, increasing the launch frequency commensal uses, or example launching probes or increasing the collector area?” to different outer planet or exoplanet explo- Our cost model incorporates the cost of launches rations, since the launch schedule has more and collector area. These are incremental costs flexibility. which strongly differ between the direct and relay configurations. The cost model does not include Receiver. Due to the need for continuous opera- other costs (such as real estate and the capital cost tion in downlink reception from probes per- of the launch beamer infrastructure) that are less forming flybys of a single target, as well as materially affected by which configuration is cho- the need to maintain pointing at the probe sen. Thus the total budgetary expenditure required trajectories to that target, generally each re- will be considerably higher than the numerical val- ceiver collector will be dedicated to a single ues listed here. We do expect that the system de- target regardless of whether a direct or relay sign and testing effort will be considerably higher 12 simpler, this would not be a meaningful compari- Table 3: Relevant cost metrics Metric Units Interpretation son. While the launch energy cost is per-probe, the Data rate b bits/sec Average rate that scientific fixed capital cost of the receiver has to be amor- data is recovered, cumulative tized across the various probes over the receiver’s over all probes Total cost d dollars/year Average rate of expenditure lifetime. That amortization depends strongly on on probe launches and the the number of probes launched during that lifetime downlink receiver amortized over time and hence the launch interval T∆. In other words, Unit cost dollars/bit Average expenditure for if probes are launched with greater frequency, then e = d/b each scientific data bit returned from the probes the ammortized cost of the receiver is lower per launch than if probes are launched at lower frequency. for the relay configuration due to its greater com- To avoid any unnecessary complexities intro- plexity, but the costs of this are difficult to quantify duced by end effects (startup of operations and end and thus are not modeled. of system life) we assume that b and d are “steady The overall conclusion is that the direct con- state” averages that overlap neither the beginning figuration is strongly advantageous from an overall nor end of operations. cost perspective, even as other cost disadvantages For both configurations, increasing b (by de- for the relay configuration are neglected. creasing T∆ or increasing A) will result in a larger d. In other words, there is inevitably a larger cost as- 7.1. Cost metrics sociated with acquiring more scientific data. Our For any particular probe and ground parame- comparison methodology chooses the same b for ters the cost metrics of interest are listed in Tbl.3. both configurations, determines how d differs be- For both configurations we assume that probes are tween them, and repeats this for different values of launched at a regular interval T∆, where T∆ will b. generally differ numerically between the two configurations. Total cost d is interpreted as the annual 7.2. Cost model budget outlay for launches and the amortized cost We utilize a simple model for cost d, of terrestrial collector area. Unit cost e, the ratio C k A d = e + a (2a) of d and b, is a measure of the cost of each unit T∆ TL of science return (one bit). When e is larger, each 1 ka unit of scientific data costs more to acquire. Sci- = Ce + ρA where ρ = . (2b) T T C entists will prefer the configuration that achieves a ∆ L e smaller e because it maximizes the science return Version Eq.(2a) includes three cost parameters associated with whatever budget is available. {Ce, ka,TL} listed in Tbl.4, and makes two assump- Both b and d are cumulative over all probes, tions regarding receiver cost. First, the dominant rather than on a per-probe basis, and each can be cost contribution to the receiver that differs be- manipulated by choosing the launch interval T∆. tween the two configurations is terrestrial aperture With regard to the terrestrial collector area A there area A, and Eq.(2) assumes that the total collector is a big distinction. For the relay configuration, area cost kaA is proportional to the area A with A is predetermined by the probe receive aperture proportionality cost factor ka. The justification is area A0/2 (adjusted to account for outages), and that for fixed building blocks (optical elements, fil- thus cannot be used to manipulate b and d. The ters, detectors, etc), the required number of such direct configuration is more complicated in that b building blocks will be proportional to A. Second, can be manipulated by the unconstrained choice of Eq.(2a) assumes straight-line depreciation over re- terrestrial collector area A as well as T∆, and the ceiver lifetime TL; that is, the capital cost is amor- tradeoff between T∆ and A strongly affects d. This tized equally over all probes launched during that ambiguity is resolved by consistently choosing the lifetime. cost-optimized combination {T∆,A} that minimizes In version Eq.(2b) the number of cost metrics is d for a given b (see §Appendix C.1). reduced to just two, {Ce, ρ}. The relative collector- Could we base cost metrics on individual probes launch cost metric ρ displays clearly the relative −1 rather than averaging across all probes? Although importance of launch frequency T∆ and collector area A in contributing to cost d. 13 12

Table 4: Cost metrics ) 2 Units Interpretation Relay is cost effective LifetimeT L=10 yr 100 yr 10 Ce dollars The incremental cost of each probe 30 yr launch (primarily energy and the cost of the probe itself) 8 2 ka dollars/m Cost of a terrestrial receiver collector per unit area Direct is cost effective 6 TL years Operational lifetime of the receiver ρ years−1-m−2 Relative measure of the launch and oto olco rai $ / m in area collector of ( cost

collector costs 10 4 Cost parameter range log = a k 2 7.3. Rate model 4 5 6 7 8

The determination of cumulative scientific rate log10 (Ce = cost of one launch in $) b differs between the two configurations. For the relay configuration, b equals Rd, which is the to- Figure 8: In a log-log plot of receive collector area cost co- tal data rate supported between two relay probes efficient ka vs the cost of each launch Ce, the dashed lines divide these cost parameters into two regions. Above the during downlink operation. For the direct config- dashed line, the relay configuration is more cost effective uration, b has to be determined in two steps. It and below that dashed line the direct configuration is more equals b = V0/T∆, where V0 is the total volume cost effective. The probe parameters in Tbl.2 are assumed, that reaches the terrestrial receiver from each in- and the bypass factor Jr=3 is adopted for the relay configuration to ensure reliable system operation. As the system dividual probe. This volume is determined from lifetime increases, the regime in which the direct configu- the initial data rate R0 and the duration of the ration is advantageous expands because the amortized cost downlink transmission (see Eq.(A.2)). Because b is of the collector area is reduced. The cost parameter range an average over all probes, the fact that individual shown adopts the values from Tbl.5, and falls well within the region wherein the direct configuration is more cost effective. probe downlink transmissions actually time-overlap one another is irrelevant to the determination of b. because larger ρ increases the relative cost of the 7.4. Numerical cost results collector area A. In the following we calculate and plot and com- For the probe and trajectory parameters in Tbl.2, −2 −1 pare total cost d and unit cost e vs average data the threshold is ρ0=2.96 m -yr . The actual value rate b for the direct and relay configurations. Al- ρ, which is listed in Tbl.5, is about five orders of though there are many design and technology el- magnitude smaller, putting the cost parameters cho- ements that will establish the cost parameters, a sen in Tbl.5 well into the regime wherein the direct generous range of possible values is listed for each configuration is more cost effective. The collector individual parameter in Tbl.5. Specific cost com- area would have to be very expensive for the re- parisons are based on specific chosen values for cost lay configuration to be cost effective; specifically 8 2 parameters. These fall within the range and are ka>5.3·10 $/m (more than half a billion dollars consistent with values used in a system study of per square meter) when TL=30 yr. the sail and propulsion system for StarShot [3]. The three cost parameters {Ce, ka,TL} contribute to ρ, and the boundary for which ρ = ρ0 is plotted 7.4.1. Is relay or direct more cost effective? in Fig.8 for three different receiver lifetimes. These The bottom line in the cost comparison is “which contours separate the parameters into two regions, is most cost effective, the direct or the relay con- one region above (within which the relay configura- figuration?” for equivalent b. This question can be tion is more cost effective) and the other region be- answered simply in terms of ρ as defined in Eq.(2b). low (within which the direct configuration is more For any set of probe and trajectory parameters as cost effective). Also illustrated by a shaded box is the range of cost parameters listed in Tbl.5. The di- in Tbl.2 there is a threshold value ρ0. For equal rect configuration is more cost effective (by a large bd = br, if ρ < ρ0 then dd < dr and ed < er, or in other words the direct configuration is more cost ef- margin) over this entire range. fective than the relay configuration (see §Appendix 7.4.2. Total cost comparison C.2). Conversely, whenever ρ > ρ0 the relay configuration is most cost effective. This is expected In the following cost comparisons between the direct and relay configurations, the chosen values 14 Table 5: Numerical cost parameters 11 Parameter range Chosen value 10 $100K ≤ Ce ≤ $10M $6M relay −2 −2 −2 $100 m ≤ ka ≤ $100K m $10K m 9 10 yr ≤ TL ≤ 100 yr 30 yr ρ 5.5 · 10−5 m−2-yr−1 8

7 direct

6 cs fiinyi dollars / bit ) in efficiency cost = ( e 12 10 log 5

relay 4 10 -2 0 2 4 6

log10(b= total data rate in bits/sec )

8 direct Figure 10: For the same conditions as Fig.9, a log-log plot ttlcs ndollars / yr ) in cost total = ( d

10 of the unit cost e in dollars/bit vs the scientific data rate b in log bits/sec. The unit cost ranges from $336M per bit to $11K 6 per bit for the direct configuration, and from $78B per bit to $2.4M per bit for the relay configuration. -2 0 2 4 6

log10(b= total data rate in bits/sec ) 6

relay Figure 9: A log-log plot of the total launch and collector hour area cost d in dollars/yr vs the total scientific data rate b 4 in bits/sec. Many other budgetary expenditures that do not day materially affect the relay vs direct configuration compari- week son are neglected. The probe and trajectory parameters are 2 those of Tbl.2, and in the relay configuration it is assumed month that Jr=3. Both d and b are aggregated over all probes year concurrently contributing to the downlink (the data volume 0 anhfeunyi probes / yr ) in frequency ( launch per probe is shown later in Fig.12). Thus d is the total an- 10 log nual budget required to fund the launches plus the amortized direct cost of the terrestrial collector area. The relay configuration -2 is consistently a factor of 239 more expensive. The annual -2 0 2 4 6 budget ranges from $335K to $10.6B for the direct configu- log (b= total data rate in bits/sec ) ration, and from $77M to $2.4T for the relay configuration. 10 Figure 11: For the same conditions as Fig.9, a log-log plot of the frequency of probe launches in probes/yr vs the scientific for cost parameters listed in Tbl.5 are adopted. The data rate b in bits/sec. Also identified are the probe launch value of ρ establishes in advance that these chosen frequencies for a launch interval equal to an hour, day, week, cost parameters strongly favor the direct configura- month, and year. The frequency ranges from 0.03 to 885 tion. launches/yr in the direct configuration, and from 13 to 409K in the relay configuration. The total cost d is shown in Fig.9 over a range of data rates b. Note that both d and b are cumulative over all probes currently operating downlinks. As 7.4.4. Probe launch frequency expected, d increases with b; that is, obtaining more Focusing on the total cost and data rate across scientific data requires a larger budget. all probes concurrently operating downlinks obscures significant differences between the direct and relay 7.4.3. Unit cost comparison configurations in launch interval, number of probes, The unit cost e is shown in Fig.10. This value is and data volume per probe. In particular, the fre- −1 the same whether determined on a per-probe basis quency of probe launches T∆ is considerably higher or cumulative over all probes. The value of e de- in the relay configuration as shown in Fig.11. This creasing with b is indicative of economies of scale. is because the unit cost e of the direct configuration That is, it is more cost effective to build one large can be reduced by launching probes less often and system rather than replicating multiple smaller sys- returning a larger volume of scientific data V0 per tems. probe.

15 tion. Its greatest advantages are its relative sim- plicity and reliability and the flexibility to achieve 10 greater data volume per probe through the uncon- direct strained choice of a larger terrestrial collector area. 8 The direct configuration is also superior on measures of annual cost (annual budgetary outlays) and relay 6 unit cost (cost per unit of scientific data) over a

aavlm e rb nbits ) in probe per volume ( data wide range of cost parameter values. 10

log This study suﬀers several limitations. By ne- 4 glecting many adverse technical factors outlined in

-2 0 2 4 6 §2.2, it presents a decidedly optimistic bound on

log10(b= total data rate in bits/sec ) the performance of the relay configuration. These issues would have to be further and more accurately Figure 12: For the same conditions as Fig.9, a log-log plot quantified before serious consideration of the relay of the per-probe data volume V0 returned to a terrestrial configuration. receiver in bits vs the total data rate b in bits/sec cumulative over all probes. This data volume is about three orders of This study has assumed a homogeneous-probe magnitude larger in the direct configuration, compensating system in which all probes are functionally equiv- for the lower launch frequency shown in Fig.11. The relay alent, both collecting and communicating scientific configuration V0 ranges from 2.4 kb to 77 Mb, and the direct configuration from 1.1 Mb to 36 Gb. data back to earth, and are launched at fixed intervals. It would be useful to study alternatives in which probes are functionally specialized, for ex- 7.4.5. Data volume per probe ample with some probes devoted to scientific obser- The data volume returned from each probe is vations and communications and others exclusively shown in Fig.12. While the number of probes mak- to communications relay. Launch schedule alterna- ing scientific observations (for equivalent b) is much tives in which probes are launched at different ve- larger in the relay configuration, the volume of sci- locities and the launch schedule is manipulated such entific data returned per probe is about three orders that probes arrive at the target in spatial groupings of magnitude smaller. This has considerable im- could also be considered. plications to the science mission, since the volume of scientific data acquired by each probe is much Acknowledgements smaller. In other words, for equivalent total data rate b the scientific observations need to be consid- PML gratefully acknowledges funding from NASA erably more fragmented among probes. NIAC NNX15AL91G and NASA NIAC NNX16AL32G It is likely that the data volume of a direct con- for the NASA Starlight program and the NASA figuration probe as assumed in Fig.12 will also be California Space Grant NASA NNX10AT93H, a gen- a mismatch to scientific needs. Fortuitously the erous gift from the Emmett and Gladys W. Tech- ground rules of this comparison (periodic probe launches nology Fund, as well as support from the Break- with launch frequency chosen for cost optimization) through Foundation for its Breakthrough StarShot can be readily violated in the direct (but not relay) program. More details on the NASA Starlight pro- configuration. For the direct configuration there is gram can be found at www.deepspace.ucsb.edu/ no constraint that launches occur at regular inter- Starlight. vals, nor is homogeneity in the mass/velocity of the various probes required. Thus there is the intrinsic flexibility to accommodate heterogenous scientific needs among the various probes in a swarm.

8. Conclusions

Although a relay configuration has some identi- fiable advantages, overall the results of this study are consistently favorable to the direct configura-

16 Appendix A. Technical model Under these assumptions the data volume V0 per probe is the same in the two cases, Appendix A.1. End-to-end metric r The numerical results in §4 and §5.2 make use A0 bP0 V0 = · . of the end-to-end power-area2 metric [1] 2(Jr + 1)u0 α Likewise the number of probes in transit is the same T T R 2 hcλ0 P A A = α0 · D R , α0 = . (A.1) in the two cases, A e e η · BPP D 2D (J + 1) rα b This relates the product of transmitted average power 0 = 0 r · 0 . T T u0T∆ A0 P0 PA , transmit aperture effective area Ae , and re- R ceive collector total effective area Ae to the prop- The data volumes per probe are calculated dif- agation distance D and rate R at which scientific ferently because the different downlinks operate con- data is recovered reliably at the receiver. The value currently in the direct configuration but not the re- of α0 is held fixed across all comparisons. We have lay configuration. For the relay configuration, the R S S S made the simplification Ae = N Ae where N is downlink operation duration equals the launch in- the number of apertures that comprise the total col- terval T∆, and thus V0 = T∆ ·Rr, where Rr is the S lector, and Ae is the effective area of an aperture. data rate between relay probes. For the direct con- R It is notable that Ae is not an effective area in the figuration, the data volume is [1] sense of antenna theory, because the receive col- R0D0 D0 lector is not a single-mode diffraction-limited aper- V0 = 1 − (A.2) S u0 D1 ture, but rather is a total of the effective areas Ae of N S such apertures. where R0 is the initial data rate following target- Eq. Eq.(A.1) is an invariant relation in the sense star encounter at D = D0. Eq.(A.2) takes into ac- that all consistent sets of parameters must satisfy it. count the declining value of R with distance D. It is simple to apply because it is not dependent on the background radiation sources and model, and the resulting signal-to-background ratio SBR at the Appendix B. Bernoulli trial statistics receive aperture. However, it will not apply to any In the language of statistics, a missing or non- arbitrary set of parameters because they may be operational probe is called a failure and a present inconsistent with one another. The principle sit- and fully operational probe is called a success. As- uation where Eq.(A.1) is invalid occurs when the sume that failures are statistically independent and signal-to-background ratio SBR at the receiver is uniformly distributed and occur with probability too small to support the assumed photon efficiency p, and define q = 1 − p. A sequence of n failures BPP for a particular background radiation model. & successes and labeled with index 1 ≤ i ≤ n is T In this event, it will be necessary to increase PA termed a Bernoulli trials sequence of length n.A sufficiently to bring SBR into line with BPP before run of k failures is defined as k consecutive failures Eq.(A.1) becomes valid. In spite of this shortcom- that are preceded and followed by successes. Note ing, Eq.(A.1) is suitable for upper bounding the that there can be multiple such runs, as long as they data rate R that can be achieved for any set of are interspersed with one or more successes. Fail- parameters, which is the goal here. ures in positions [1, k] followed by a success, as well The terrestrial collector areas are determined as in positions [n − k + 1, n] preceded by a success, from Eq.(A.1), substituting values from Tbl.1 and also qualify as runs. setting equal data rates br = bd = b and equal launch Define L as the length of the longest run of fail- intervals T∆ in the two cases, ures in a Bernoulli trials sequence with total length q 2 n. L is a random variable with cumulative distri- 1 P0 2α0b A0 + Dl bution function given by 2 α0b An = A0P0(1 − Po) Pr {L ≤ k − 1} = α bD D 0 0 1 b n+1 c Ad = √ . k+1 2(J + 1)(1 − P )(D − D ) α bP X n − mk n − mk r o 1 0 0 0 (−1)mpmkqm−1 + q m − 1 m m=0 17 for k ≥ 1 [9]. Appendix C.2. When the relay configuration is cost In the relay configuration, all runs of length effective k ≤ J can be tolerated without relay system fail- r Although the expression for dr in Eq.(C.1b) is ure, where Jr is defined in §5.2. Relay system fail- complex, it simplifies considerably when ρ=0. Since ure results whenever L > Jr, and hence the proba- A is very small in the relay configuration in any bility of system failure is case (it is constrained by the probe receive aperture, and hence the probe mass), arbitrarily setting Pr {system failure} = ρ=0 should not affect dr materially. Thus to find Pr {L > Jr} = 1 − Pr {L ≤ Jr} the region where the relay configuration has lower cost, it is a good approximation to use the modified for J ≥0. r criterion

dr ρ=0 < dd . (C.3) Appendix C. System cost results This criterion slightly expands the region wherein The cost model is given in Eq.(2). This can the relay configuration is more cost effective, since be combined with Eq.(A.1) and Eq.(A.2) to yield it deliberately ignores the cost associated with A. formulas for cost metrics b, d, and e as a function of Thus, the “relay is cost effective” region in Fig.8 is probe and probe trajectory parameters. Expressed actually slightly smaller than shown. The critical in terms of rates bd and br, these are value of ρ where equality is achieved in Eq.(C.3) is

2 dd 1 α0bdD0D1ρT∆u0 (Jr + 1) (1 − Po)u0(D1 − D0) = + (C.1a) ρ0 = . (C.4) Ce T∆ A0P0(1 − Po)(D1 − D0) A D D √ 0 0 1 d 2α b D2ρ 2(J + 1)u α b P r = 0 r l + r 0 0 r 0 Thus, the relay option becomes more cost-eﬀective Ce A0P0(1 − Po) A0P0 √ when probe aperture area A0 is larger or probe by- A0ρ 2Dlρ α0brP0 pass parameter J is smaller. + + . (C.1b) r 2(1 − Po) P0(1 − Po) References

Appendix C.1. Direct conﬁguration cost optimiza- [1] D. G. Messerschmitt, P.Lubin, I. Morrison, Challenges in tion scientiﬁc data communication from low-mass interstellar probe, arXiv preprint arXiv:1801.07778 (2020). While dr in Eq.(C.1b) is not dependent on T∆ [2] D. Messerschmitt, P. Lubin, I. Morrison, Technological (because this parameter was manipulated to achieve challenges in low-mass interstellar probe communication, rate b ), d in Eq.(C.1a) is dependent on T (be- arXiv preprint arXiv:2001.09987 (November 2019). r d ∆ [3] K. Parkin, A starshot communication downlink, arXiv cause Ad rather than T∆ was manipulated to achieve preprint arXiv:2005.08940 (October 2019). bd). There is therefore an opportunity to minimize [4] B. Oliver, J. Pierce, C. E. Shannon, The philosophy of pcm, Proceedings of the IRE 36 (11) (1948) 1324–1331. dd by choosing the cost-optimum value of T∆.A restatement is that the two performance metrics [5] J. Gramling, N. Chrissotimos, Three generations of nasa’s tracking and data relay satellite system, in: {bd, dd} depend on the two parameters {T∆,A}, SpaceOps 2008 Conference, 2008, p. 3313. and there is thus an opportunity to choose the com- [6] K. Bhasin, A. Hackenberg, R. Slywczak, P. Bose, M. Bergamo, J. Hayden, Lunar relay satellite network for bination that both achieves rate bd and at the same space exploration: architecture, technologies and chal- time minimizes dd. The resulting values of T∆ and lenges, in: 24th AIAA International Communications dd are Satellite Systems Conference, 2006, p. 5363. [7] R. Hastrup, R. Cesarone, A. Miller, R. McOmber, Mars 2 A0P0(1 − Po)(D1 − Do) relay satellite: Key to enabling & enhancing low cost T∆ = (C.2a) exploration missions, Acta Astronautica 35 (1995) 367– α0bdD0D1ρu0 s 376. dd α0bdD0D1ρu0 [8] M. Matsumoto, Fiber-based all-optical signal regenera- = 2 . (C.2b) tion, IEEE Journal of Selected Topics in Quantum Elec- Ce A0P0(1 − Po)(D1 − D0) tronics 18 (2) (2011) 738–752. [9] M. Muselli, Simple expressions for success run distribu- The conclusion√ is that dd in Eq.(C.2b) is propor- tions in bernoulli trials, Statistics & Probability Letters tional to bd, while dr in Eq.(C.1b) has a more 31 (2) (1996) 121–128. doi:10.1016/S0167-7152(96) 00022-3. complicated dependence on br. 18