The advantages of neutral particle beam propulsion seem clear: Whereas a laser’s photon beams can exchange momentum with the sail, neutral particle beams transfer energy and are considerably more efficient. In fact, as we saw in the first part of this essay, that efficiency can approach 100 percent. A mission concept emerges, one that reaches a nearby star in a matter of decades. But what about the particle beam generators themselves, and the hard engineering issues that demand solution? For that matter, how does the concept compare with Breakthrough Starshot? Read on as James Benford, working in collaboration with Alan Mole, describes the salient issues involved in building an interstellar infrastructure.

By James Benford and Alan Mole

We discuss the concept for a 1 kg probe that can be sent to a nearby star in about seventy years using neutral beam propulsion and a magnetic sail. We describe key elements of neutral particle beam generators, their engineering issues, cost structure and practical realities. Comparison with the Starshot laser beam-driven concept gives roughly similar costs.

Beam Generator Concept

Figure 1 Block diagram of early neutral particle beam generator [1]. Drift-Tube Linac is not shown.

Creation of the neutral particle beam begins with

1. Extraction of a negative ion beam (negative ion with attached electrons) from a plasma source; it then drifts into the first acceleration stage, the RFQ. The first element of the accelerator will appear much like the geometry shown in figure 2. Here ions are extracted from the plasma source on the left by electrostatics and brought by a converging magnetic field to the linear accelerator.

Figure 2. Ion beam on left is propagated along converging magnetic field to the linac.

2. The ion beam enters a radiofrequency quadrupole (RFQ) accelerator, a vane-like structure where the application of radiofrequency power produces a continuous gentle acceleration much like a surfer riding a wave. It also provides strong electrostatic focusing to prevent divergence growth. The structure bunches the particles in phase space.

The RFQ fulfils at the same time three different functions:

  • focusing of the particle beam by an electric quadrupole field, particularly valuable at low energy where space charge forces are strong and conventional magnetic quadrupoles are less effective;
  • adiabatic bunching of the beam: starting from the continuous beam produced by the source it creates with minimum beam loss the bunches at the basic RF frequency that are required for acceleration in the subsequent structures;
  • acceleration of the beam from the extraction energy of the source to the minimum required for injection into the following linac structure.

3. After the ions exit the RFQ at energies of a few MeV, further acceleration to increase the particle energy is done with a drift-tube linac (DTL), which consists of drift tubes separated by acceleration regions, as shown in Figure 3. Particles arriving at the gaps at the proper phase in the radiofrequency waves are given acceleration impulses. When the electric field of the wave reverses, the particles are shielded from being accelerated by passing through the drift tubes. The typical accelerating gradient is a few MeV/m.

Figure 3. Drift-Tube Linac, which consists of drift tubes separated by acceleration regions.

4. In order to maintain low emittance and produce the microradian divergence we desire, the beam is expanded considerably as it exits the accelerator. Beam handling elements must have minimal chromatic and spherical aberrations.

5. Beam pointing to be done by bending magnets with large apertures.

6. Finally, the extra electrons are stripped from the beam, making a neutral particle beam. This can be done with by stripping the electrons in a gas neutralization cell or by photodetachment with a laser beam. It may be possible to achieve 100% neutralization by a combination of methods. Thus far this high-efficiency neutralization has not been demonstrated.

Beamer Engineering

There are several possible schemes for building the beam generator. Both electrostatic and electromagnetic accelerators have been developed to produce high power beams. The most likely approach is to use linear accelerators. In the past, the cost of an electromagnetic accelerator is on the order of a person year per meter of accelerator (~1 man-year/m) but this could be larger for the more sophisticated technologies.

The power system to drive such accelerators could come from nuclear power (fission or fusion) or solar power. Furthermore, if it were to be space-based, the heavy mass of the TW-level high average power required would mean a substantially massive system in orbit. Therefore Mole’s suggestion, that the neutral beam be sited on Earth, has its attractions. There is also the question of the effects of propagating in the atmosphere, on both beam attenuation and on divergence.
If the beam generator were to be on Earth, it should all be at the highest altitude for practical operations. The Atacama Desert, for example, would offer very low humidity and half of sea level pressure. In addition, a way to reduce beam losses in the atmosphere would be to launch a hole-boring laser beam in advance just before the neutral beam. This laser would heat up a cylinder of atmosphere to lower the pressure, allowing the neutral beam to propagate with less loss. Such hole-boring exercises have been conducted in laser weapon studies and does appear to be a viable technique.

The final neutral beam can be generated by many small beam drivers or a single large beam driver. If a great number of driver devices and their associated power supplies are required, increasing the construction and maintenance expense of this portion. Of course, economies of scale will reduce the cost of individual segments of the Beamer by mass production of the system modules. Making such choices is an exercise for future engineers and designers.

Neutral particle beam generators so far have been operated in pulsed mode of at most a microsecond with pulse power equipment at high voltage. Going to continuous beams, which would be necessary for the seconds of beam operation that are required as a minimum for useful missions, would require rethinking the construction and operation of the generator. The average power requirement is quite high, and any adequate cost estimate would have to include substantial prime power and pulsed power (voltage multiplication) equipment, the major cost element in the system. Of course, it will vastly exceed the cost of the Magsails, which is an economic advantage of beamed propulsion.

However, this needs economic analysis to see what the cost optimum would actually be. Such analysis would take into account the economies of scale of a large system as well as the cost to launch into space versus the advantages of beaming from Earth.

Beamer Cost Estimates

The interstellar neutral particle beam system described here is a substantial extrapolation beyond the present state-of-the-art. Nevertheless, estimates can be made of both the capital and operating costs.

The cost of the Beamer is divided between the cost of the accelerator structure (RFQ and DTL) and the power system that drives it. For a cost estimate for the Mercury system, we assume that the present day accelerating gradient is maintained for this very high-power system. That gradient is ~ 2 MeV/m. For the mercury neutral particle beam the length of the 1.35 GeV accelerator would be 675 m.

There is an extensive technology base for drift-tube linacs; there are many in operation around the world [2]. We use as a model the well-documented 200 MeV Brooklyn National Laboratory 200 MeV ion beam system, which was completed in 1978 at a cost of $47M. It used 22 MW of radiofrequency power and was 145m long. In that era, the cost of microwave equipment was ~$1/W. The cost today is ~$3/W, so the 22 MW would cost 22 M$ then and 66 M$ today. Since the total cost of accelerator was $47 M$, the Accelerator structure would cost 47 M$ -22 M$ = $25 M$. Thus at this level the two cost elements are roughly equal. The accelerator structure then costs $25 M$/145 m = $0.17 M$ per meter in 1978. We multiply all costs by a factor of three to account for inflation to get today’s costs.

To estimate the capital cost of the mercury in NPB described here, we have the following relations:

Caccl= 0.5 M$/m x 675 m = 350 M$

Cmicrowave= 3$/W x 18 TW = 5.47 B$

Therefore the dominant cost element would be the microwave system driving the accelerator.

However, high-volume manufacturing will drive costs down. Such economies of scale are accounted for by the learning curve, the decrease in unit cost of hardware with increasing production. This is expressed as the cost reduction for each doubling of the number of units, the learning curve factor f. This factor typically varies with differing fractions of labor and automation, 0.7 < f < 1, the latter value being total automation.

It is well documented that microwave sources have an 85% learning curve, f = 0.85 based on large-scale production of antennas, magnetrons, klystrons, etc [3]. Today’s cost is about $3/W for ~1 MW systems. Note that this includes not only the microwave generating tube, but also the power system to drive that continuous power. The 18 TW power needed would require 18 million such units. Therefore the cost is ~1.1 B$. Adding together the accelerator and microwave power system, the cost will be 1.45 B$.

The electrical power to drive this large system cannot possibly come from the electrical grid of Earth. Therefore a large cost element will be the system that stores the 162 TJ of energy. (Note that the beam power starts at zero and rises with time (t2) to 18 TW at the end.) From Parkin’s estimates of the Starshot energy storage system [10], based on Li-ion batteries, we take the storage cost to be $50 per kilowatt-hour, which is $13,900 $/TJ. Consequently the cost for the energy store is ($13,900 $/TJ) 162 TJ = 2.25 B$. So the energy stores cost is comparable to that of the accelerator.

The total capital cost is

Caccl= 350 M$

Cmicrowave = 1.1 B$

Cstore= 2.25 B$

Total accelerator capital cost is 3.7 B$.

The operating cost to launch a single Magsail is of course far smaller. It is simply the cost of the spacecraft and the energy to launch it. We will assume that the cost of the spacecraft will be on the order of $10 million. The cost of the electricity at the current rate of $.10 per kilowatt-hour is $4.5 million.

Total operating cost for a single launch is ~15M$.

Comparison with Starshot

The neutral particle beam approach is conceptually similar to photon beams such as the laser-driven Starshot project. A disadvantage of reflecting photons from the sail will be that they carry away much of the energy because they exchange only momentum with the sail. Neutral particle beams transfer energy, which is much more efficient. The reflecting particles may in principle be left on moving in space after reflection and thus the efficient energy efficiency can approach 100%.

The Starshot system, a laser beam-driven 1 gram sail with the goal of reaching 0.2c, has been quantified in a detailed system model by Kevin Parkin [4]. Since both the high acceleration neutral particle beam described here and Starshot are both beam-driven high-velocity systems, we make the following comparison between their key parameters and cost elements:

Physical parameters and cost elements of beam-driven probes

Mercury Neutral Particle Beam SystemStarshot
Sail mass1 kg1 g
Velocity0.06 c0.2 c
Beamer capital cost1.45 B$4.9 B$
Energy store cost2.25 B$3.4 B$
Total capital cost3.7 B$8.3 B$
Energy cost/launch4.5 M$7 M$
Kinetic energy1.6 1014 J1.8 1012 J
Kinetic energy/ capital cost43.2 kJ/$0.2 kJ/$

Here we have summed the accelerator and microwave power system costs for the neutral Beamer and the laser and optics cost for Starshot. A major caveat is that Parkin’s estimates have realistic efficiencies of the systems of Starshot, but our costs assume unrealistically high efficiencies.

Although they differ in detail, the two concepts give the same order of magnitude cost. However, the kinetic energy in the NPB-driven probe is 90 times that of the Starshot probe. This shows the disadvantage of reflecting photons from the sail: they carry away much of the energy because they exchange only momentum with the sail. Neutral particle beams transfer energy, which is much more efficient. The kinetic energy/capital cost ratio is 200 times greater in the NPB case.

It is instructive that the high-energy requirement of interstellar probes drives the existence of a stand–alone storage system, which is a major element in the total cost of both systems. The similarity of costs for these rather different beam- driven systems gives us some confidence that these rough estimates in this paper are credible.

Neutral Particle Beam Realities

Practical realities are always bad news. Performance of most systems degrades to below their design points because of inefficiencies of processes. Note that the beam systems described here are perfectly efficient, as determined from equation 5. That is, the beam reflects from the sailcraft with perfect efficiency, so as to stop dead, transferring all the energy to the spacecraft. The realities of neutral particle beams in the present day are substantially poorer.

To see where the problems lie, we consider a daring experiment called BEAR, conducted 30 years ago [1, 5]. A neutral particle beam generator was actually deployed and operated in space and its performance was measured.

On July 13, 1989 the Beam Experiment Aboard Rocket (BEAR) linear accelerator was successfully launched and operated in space by Los Alamos National Laborotory. The rocket trajectory was sub-orbital, reaching altitude of 220 km. The flight demonstrated that a neutral hydrogen beam could be successfully propagated in an exoatmospheric environment. The cross-section of the rocket is shown in figure 4.

Figure 4. Beam Experiment Aboard Rocket (BEAR) [1].

The accelerator, which was the result of an extensive collaboration between Los Alamos National Laboratory and industrial partners, was designed to produce a 10 rnA, 1 MeV neutral hydrogen beam in 50 microsecond pulses at 5 Hz. The major components were a 30 kev H- injector a 1 MeV radio frequency quadrupole, two 425 MHz RF amplifiers, a gas cell neutralizer, beam optics, vacuum system and controls. The beam extracted was 1 cm in diameter with a beam divergence of 1 mradian. There was no unexpected behavior such as beam instability in space.

The design was strongly constrained by the need for a light- weight rugged system that would survive the rigors of launch and operate autonomously. The payload was parachuted back to Earth. Following the flight the accelerator was recovered and successfully operated again in the laboratory.

From the paper and report describing this experiment we see substantial inefficiencies, which should guide our future expectations.

The input power to the accelerator was 620 kW for 60 µs, a 7.2 J energy input. The beam as extracted was 27 mA at 1 MeV for 50 µs, which gives 1.35 J. The efficiency therefore is 19%, so approximately 4/5 of the energy supplied was lost in the beamline shown in figure 1. The major loss was in the neutralizer which was a xenon gas injected into the beamline. The efficiency of the neutralizer was changed by varying the amount of gas injected. They obtained 50% neutral hydrogen and 25% each of negative and positive hydrogen. Therefore the neutralization process was only 50% efficient in producing a neutral beam. This accounts for most of the loss. The other losses can be accounted for by inefficiencies in the optics of the low-energy beam region and the high-energy beam region.

In the 30 years since the flight, little work on particle beams has occurred at high power levels, because of the termination of the Strategic Defense Initiative. Doubtless substantial improvements can be made in the efficiency of NPB’s, given substantial research funding. Therefore the concept in this paper, with its hundred percent efficiency of energy transfer from the electrical system to the sail, is an upper bound on the performance. Consequently the parameters in Table 1 and the capital and operating cost estimates given here are lower bounds on what would actually occur.


The cost model presented here is lacking in realistic efficiencies. The next level of analysis should address this lack.

We can forsee a development path: a System starts with lower speed, lower mass Magsails for faster missions in the inner solar system. As the system grows, the neutral beam System grows and technology improves. Economies of scale lead to faster missions with larger payloads. As interplanetary commerce begins to develop, making commerce operate efficiently, outcompeting the long transit times of rockets between the planets and asteroids, the System evolves [6]. Nordley and Crowl describe such a development scenario [7]. We conclude that this concept is a promising method for interstellar travel.


1. P. G. Oshey, T. A. Butler, M. T. Lynch, K. F. McKenna, M. B. Pongratz, T. J. Zaugg, “A Linear Accelerator In Space-The Beam Experiment Aboard Rocket”, Proceedings of the Linear Accelerator Conference 1990.

2. H. B. Knowles, “Thirty-Five Years of Drift-Tube Linac Experience” Los Alamos Scientific Laboratory Report, LA-10138-MS, 1984. See also reference 4, pg. 81.

3. J. Benford, J. A. Swegle and E. Schamiloglu, High Power Microwaves, Third Edition, pg. 77, Taylor and Francis, Boca Raton, FL, (2015).

4. K. L. G. Parkin, “The Breakthrough Starshot System Model”, Acta Astronautica 152, 370-384, 2018.

5. G. J. Nutz, “Beam Experiments Aboard a Rocket (BEAR) Project Summary’, LA-11737, 1990.

6. J, Benford, “Beam-Driven Sails and Divergence of Neutral Particle Beams” JBIS 70, pg. 449-452, 2017.

7. G. Nordley and A. J. Crowl, “Mass Beam Propulsion, An Overview”, JBIS 68, pp. 153-166, 2015.