I have a couple of slight corrections to Francis Vaughn’s otherwise excellent summary: Liquid propellant systems are referred to as engines; only solid propulsion systems are called motors. The other is the claim that the addition of deployable legs do not add to the technical complexity; in fact, developing landing legs which can be reliability deployed in such a way as to securely land the vehicle but not contribute an egregious amount of extra weight turns out to be a very complicated problem, and one that I’m not at all sure SpaceX will make work reliably. The legs have to be secure enough to hold the vehicle against wind loads and structural oscillation but compliant enough to allow some degree of off-axis landing without inducing excessive offset load.
The rationale for powered landing versus the use of aerodynamic decelerators (parachutes, ballutes, et cetera) has three components; one is that the engines are an existing system while adding a deployable decelerator adds another system with attendant dead mass. The second is that decelerators require repacking, and unlike personnel chutes, these are difficult and complex to pack, requiring complex folding and a heavy press. The third is that parachutes and ballutes are single shot textile devices that cannot be functionally verified before use with no redundancy or requiring significant excess mass to have redundant systems.
This gets back to the question of the essential value of reuse; e.g. is there a cost or flight frequency rationale for trying to reuse flight hardware? From a naive standpoint reuse seems like an essential component of reducing launch costs, with advocates comparing current single use systems to flying an airliner once and then throwing it away. However, this ignores the fact that the usable lifetime of airliner engines is hundreds of hours between major servicing and thousands of hours between rebuild or replacement, and airframes have seen tens of thousands of hours of service between repair and on the order of a hundred of thousand hours of lifetime service. The most robust rocket engines have demonstrated a few thousands of seconds of use between major rebuilds (RS-25 SSME) and only after significant development. Other pressurized components such as valves, composite overwrapped pressure vessels, and feed systems also have limited lifetimes. Other structural items on rockets see significant structural fatigue during flight, experiencing vibration levels that are orders of magnitude beyond what is seen in any normal terrestrial environment, and because the philosophy behind most rocket launch systems is to try to get the highest possible performance per mass, a key metric is reducing ‘dead’ weight (e.g. anything that is not propellant or payload) as much as possible, which emphasizes performance over robustness.
The fact of the matter is that the actual cost of building hardware, while significant, is not the major driver in total launch cost. It’s difficult to get a good handle on hardware costs because of the way costs are tracked, i.e. production costs are not clearly separated from processing and non-recurring engineering (NRE), but the studies that I’ve worked on indicate that the actual build cost of a two or three stage liquid propulsion rocket averages somewhat less than 10% of the total launch cost and never exceeds 20%. This means that, all other costs being equal and assuming no remanufacture or rebuild outside of normal processing, you’d have to fly between five and ten flights just to break even on hardware cost. And all other costs are not equal; reusability demands substantially more NRE and inspection of post-flight items, and to date nobody has built a reusable launch system of significant capability which can just land, refuel, and fly again without substantial refurbishment. The study that NASA performed in the 'Seventies indicated that a reusable multistage vehicle would be cheaper than an expendable system only at a flight rate of 50-60 vehicles per annum. Orbital Sciences redid that study in the 'Nineties and reached the exact same conclusion, which is unsurprising because engine performance and vehicle mass ratios have only improved in tiny increments, being limited by the basic physics of combustion and materials.
There may be other benefits to reusability, such as increasing the number of flights without increasing the throughput on your production line, but fabrication cost is actually not a significant driver or a major opportunity to reduce launch costs by the order-of-magnitude necessary to substantially increase access to space. The real opportunities are reduction of complexity of launch vehicles, e.g. trading performance for simplicity and robustness (which actually gains more in practical reliability over redundancy) and reducing processing and integration effort by automation and simplicity. The Sea Dragon/SEALAR concept is the prime example of this; gaining partial reusability by trading performance for a low tolerance, highly robust design that requires minimal launch infrastructure and allows for launching payloads that would be almost impossible to manage with land-based logistics. However, it is really only advantageous for large payloads that can either be securely encapsulated or tolerant against marine conditions.
The other tact–and the one I personally favor–is to abandon the traditional high L/D cylindrical profile for a launch vehicle (which is inherently structurally weak and is has poor mass utilization) to a squat profile. While this would require a very different manufacturing flow than conventional launch vehicles and is logistically complex for transportation to a manufacturing site to launch facility, it offers a better trade in mass utilization, performance, robustness, and the potential for genuine reusability with a base entry high drag reentry profile with only a modest aerodynamic penalty for an orbital ascent profile. The Chrysler Aerospace SERV proposal for the STS (which was rejected out of hand, ostensibly for not meeting the adjusted crossrange requirement but really because it was just too far away from the winged spaceplane concepts that the ‘Huntsville group’ had been pursuing since before Apollo) is the epitome of this, most most plausible RSSTO concepts and the ‘Big Dumb Booster’ type two stage proposals have relied on this kind vehicle layout.
Trying to land boosters with a horizontal glide mode requires the addition of inert mass of deployable wings (or wings that protrude during flight, increasing drag during ascent), landing systems, and reinforcing structure. Every time a proposal for liquid flyback boosters came out to replace the Shuttle SRBs, the resulting concept was nearly as large as the External Tank in order to achieve the required thrust and flight time. Powered vertical landing, if it can be done successfully, requires only a very modest amount of propellant (somewhere around 5% of the total load out), although that last 5% of propellant is a loss of somewhere around 20%-30% of ascent payload capability, so there is a significant tradeoff in terms of capability limits versus reusability, which again, is not likely to be a significant cost reduction in overall launch cost.
Stranger