Jeffrey LewisMauger on China’s Hypersonic Test

AS James Acton, Catherine Dill and I prepared our “Crashing Tiger, Hidden Hotspring” post, one of my students, Philippe Mauger, made a number of important observations including offering a possible identification of the rocket engine found among the debris. I asked Philippe to write up some of his observations.

Hypersonic loose ends
A short addendum to the “Crashing Tiger, Hidden Hotspring” piece.

Philippe Mauger

Identifying Taiyuan as the correct launch site

A publicly announced satellite launch took place at Jiuquan at 1:45pm on August 9, less than two days after the crash. This fact should have immediately raised doubts about the initial reports that Jiuquan had been used as the launch site.

This local picture of the launch in Taiyuan, which shows the sun’s rays (top left) and a rocket plume (heading right, perpendicular to the rays), provides an additional hint that the launch was part of a military test. SunCalc can be used to fix the position of the sun in Taiyuan at the time of the test– but suffice to say that the East is, indeed, red. The rocket is thus westbound. As the main piece notes, nearly all satellite launches instead use an eastwards trajectory to take advantage of the Earth’s rotation.

Liquid fueled two-stage rockets: drawing from the Soviet experience

As analyzed in the main piece, the debris markings and the fuel signature are consistent with–if not foolproof evidence of– the use of a China Aerospace (中国航天) Long March (LM) rocket. Although it is difficult to tell whether a first or second stage crashed (more on this below), it is reasonable to rule out a three-stage rocket, which would be overkill given the required performance requirements. Details about a Soviet/Russian hypersonic weapons program, published on Pavel Podvig’s blog, provide a useful reference point. It supports the two-stage-only claim, given that the Russians apparently use the two-stage, liquid-fueled SS-19 (UR-100NUTTH) as their carrier rocket.

Page 2 of the LM document referenced in the main piece provides characteristics for the entire LM rocket family. SS-19 characteristics, taken from Russia’s Arms and Technologies, Volume I: Strategic Nuclear Forces (p.69, 77-79), are reproduced below. UR-100N base characteristics are also included to fill in some gaps; these values are given in an older catalog: Russian Armament State Corporation’s Russia’s Arms Catalog, Volume VI: Missiles and Space Technology (p. 268). The two-stage Chinese LM-2C is a larger rocket than the SS-19 ICBM. Also note that the payload capacity values should not be blindly compared, given that rocket performance is highly dependent on the firing mode and mission range, which differs for both missiles. In all, even when allowing a conservative performance margin to account for a heavier and/or larger Chinese hypersonic warhead, the use of a three-stage LM rocket appears uncessary.

Characteristic Russian UR-100N Russian SS-19 Chinese LM-2C
Number of stages 2 2 2
Solid or liquid fuel(s)? Both stages liquid Both stages liquid Both stages liquid
Height (m) 24 43.0
Diameter (m) 2.5 3.35
Launch weight (T) 103 105.6 245
Lift-off thrust (kN) 2962
Payload capacity (kg) 4350 (ballistic, 10,000km range) 3850 (space launch to LEO)

The engine in the debris field, and estimating the HGV’s test range

Given the above, the crash photographs are believed to be of a first or second LM stage. Identifying which stage, however, is nontrivial. One good quality picture of the crash debris shows a badly damaged engine. In its pictured state, figuring out which part used to go where is a 10,000-piece, 3D, puzzle. But while the identification is not definitive, the rocket engine does not appear to be part of a YF-21 cluster (a cluster of four YF-20 engines) used in LM first stages. The argument in support of this claim is two-fold.

First, according to a reference manual (section 6-7: Republic of China) hosted on the FAS website, an LM first stage should have a cluster of four YF-20 engines. According to the same document, an LM second stage has a single, different, engine: a YF-22 or a YF-25. The debris, despite begin densely packed, shows the remains of only one engine. It is possible that three other engines were scattered farther out in the crash, but this is less likely than the single-engine hypothesis.

Second, each YF-20 engine in a YF-21 cluster has a “pinched” zone at the top, which does not seem to appear on the engine pictured. The website maintains a wonderful collection of rocket engine pictures for comparison, including one of a YF-21 cluster, a YF-22, and an extended YF-22 (YF-22E). The following composite image shows the relevant portion of the broken engine in the debris field, boxed out; the “pinched” zone on a YF-20 engine taken from the linked YF-21 cluster picture; and a standard YF-22 for comparison.

If the photographs do indeed show a second stage, what follows? Given that this presumed last stage still had fuel remaining and thus did not reach its designated range, it is reasonable to conclude that the NOTAM eastern keep-out zone was its planned drop zone. The warhead would thus have been expected to separate before the NOTAM eastern keep-out zone. The Chinese HGV would then have had an expected minimum test range somewhat in excess of 1100km (the distance between both keep-out zones) and well below 1750km (the Taiyuan-Western keep-out zone distance).


  1. RAJ47 (History)

    There is a slight dichotomy here. If the pics were indeed of a second stage and 中国航天 (China Aerospace) written on it, logically implies that it was a three stage rocket. The characters 中国航天 are generally written on the first stage of a two stage rocket and on the second stage of a three stage rocket.

  2. Philippe (History)

    Hi RAJ47:

    Thank you for the comment. The document linked with the characteristics for the LM rocket series classifies the LM-2F as a 2-stage rocket, and this picture of an LM-2F shows the characters on the second stage:–XwRZM3bc7k/ULuUtvUXKZI/AAAAAAAAAmY/sRHBQ–_i9g/s1600/LM-2F_Launch_Vehicle.jpg

    The picture was taken from this source:

    Based on this, I think it is possible to have the characters on the second stage of a two-stage rocket.

    The characteristics for the LM-2F are far better than those of the LM-2C (listed for reference above). I displayed the characteristics for the LM-2C because it had the most conservative values (least powerful) of the entire LM 2-stage family still in service.

  3. kktt (History)

    Your student’s identification is wrong.

    1, The latest YF-21 has no ‘pinched’ part. See pic at the website of Xi’an Aerospace Propulsion Institute, who produced YF-21:

    2, The supporting metal framework on top of YF-21 an YF-22 are differrent:
    The framework in the debris obviously has the feature of YF-21, not YF-22.

    • George William Herbert (History)

      The identification by kktt seems better; the mounting frame hardware seems pretty unambiguous. I suspect that there are more of the propellant feed lines in the picture than we have motors, which would seem to indicate some motors either are obliterated, buried under the earth, or bounced downrange some more from this debris pile.

      This would then seem to be a first stage, with all the trajectory implications etc etc.

      In terms of simplifying our choice of booster overall model ID, this doesn’t help much, as it could be a DF-5 missile or any number of CZ-2, 3, or 4 models…

      So, old surplus or new build?…?

      I have no time tonight, but someone might want to see what burnout would be on a lightly loaded DF-5 first stage or CZ-2, 2C, 3, 4 first stage with a notional HWS on top.

    • Philippe (History)

      Hi kktt:

      Thank you for your comments.

      First, I would like to re-state that I am not claiming to have identified a YF-22. I am arguing that the crashed engine does not appear to be a YF-21. This is based on:
      a) the lack of any other engine at the crash, despite the tight debris cluster; and
      b) what I claim are differences between the YF-21 and the crashed remains.

      I’ll respond to both arguments you brought up on point b):

      1) On both the YF-21 picture I linked, and the other type YF-21 pictures, there is a pronounced, raised,
      top portion. For instance, in this picture:
      The YF-21 model still has a pronounced upper lip, so to speak, at the top.
      On the debris crash picture, the top of the engine is quite flat near the end.
      Based on this, I still do not think that the crashed engine is a YF-21, although there is much
      room for disagreement.

      2) On the engine superstructure:
      The engine in the crash picture is clearly not a YF-22E (extended). You can see this from the bottom ring:
      On a standard YF-22 (like on the YF-20), it is close to the bottom:

      On a YF-22E, it is much higher:

      And I do not think it is correct to compare the top structures of the YF-21 and YF-22 using a YF-22E picture to stand in for the YF-22. In particular, the YF-22 has a stronger-looking build overall than the YF-22E; it would not outlandish for the superstructure to also be thicker.

      I think it is also worth pointing out that the engine superstructure is still nicely centered on the crashed engine, but that there is still only 1 engine visible.

    • George William Herbert (History)

      Phillipe, the structure is not centered on that engine. The structure is (or is nearly identically configured to) the square thrust frame plus four corner brackets to the tank from the linked -21 pic. One side of the square’s tubing was ripped off and is in the left foreground. Two of the angle brackets at the corners are visible to the rear corners. And the engine rotated 90 degrees (in plane of picture, clockwise) around its attach point to the square thrust frame.

    • Philippe (History)

      Ah, now I see which part you mean; I was too focused on the triangular superstructure. I stand corrected and agree the square mounting frame is definitely not in line with the engine.

      If I may ask: what do you make of the debris in the top-right, background? Is that additional engine tubing?
      (visible in the original crash picture:

    • John Schilling (History)

      OK, a Long March 2C first stage carrying an X-51 clone would have a horizontal burnout velocitty of about 6.3 km/s, which seems implausibly high. It would also have a burnout acceleration of 27G, which I think is right out. In order to keep the acceleration below 10G, the WU-14 would have to come in at 23 tonnes, or 12 tonnes for a 15G burnout acceleration. So I’m not buying this as a DF-5 or LM-2/3/4 first stage.

      A DF-3, or a DF-4 first stage, might be plausible. You’d only need a 3.3-tonne payload (and/or ballast) to keep the burnout acceleration down to 15G, with a burnout velocity of 4.5 km/s. The PRC is probably getting rid of surplus DF-3/4s as they bring the DF-21/31 on line, and the YF-2 engine with quad mounting frame looks fairly similar to the YF-20/21 on the DF-5/LM-2.

      With both stages of a DF-4 and a notional X-51 clone payload, burnout velocity would be about 6.0 km/s. However, the first stage would burnout at less than 2 km/s, so we couldn’t be seeing its wreckage 600 km downrange.

      The other possibility, and it’s a bit of a long shot, is a booster built around one the liquid strap-on boosters of the Long March 3 series. Those are in ongoing series production, making it easy to maintain a steady supply for an R&D (or even operational) hypersonics program. Burnout velocity with an X-51 clone would be about 4.4 km/s, which seems about right.

      But those boosters don’t have any sort of guidance capability, and with fixed-nozzle engines it would be a major modification to turn them into standalone vehicles. The engine would be a single YF-20 in a non-standard, but presumably symmetrical, mount.

      So, my guess would be a surplus DF-3.

    • George William Herbert (History)

      The debris in the top right background… Wow. Too few pixels to be sure, but suggestive of the propellant feed lines we see elsewhere. Has smooth section then section with what could be additional rings on it, like we see in foreground tubing.

      But, if that’s the best resolution of that which exists, we might well not be able to tell for sure from that few pixels.

    • George William Herbert (History)

      John, looking at details, two things re a DF-3 YF-2A as what’s in the crash vs a YF-21 cluster…

      One, on the brackets, there are vertical reinforcing plates between the risers and the upper square on the debris, and not on the YF-2A photos like:

      I think I see them on the YF-21 like:

      They stick a bit further down than the plates between the riser tubes. You can clearly see that in the debris and I think I see it in “engine-family_big.jpg”

      The YF-2A engine quad does have a similar brace structure but does not have the outer brace plates:

      Two, the debris clearly has some sort of regen propellant flow mainfold around the throat region (not sure above or on, but right near). YF-2 shows clear throat, no regen manifold.

    • Philippe (History)

      If it helps, this is the largest version of the engine debris I can find:

    • John Schilling (History)

      If this is DF-5/LM-2 first stage wreckage we are looking at, then I think there pretty much has to be a second stage as well. First, because as noted the burnout acceleration would be impossibly high without it (or without 10-20 tonnes of ballast). Second, because the second stage is where all the avionics are for; the first stage is dumb, controlled by the second.

      And that has some serious implications, because burnout velocity for a DF-5/LM-2C stack with an X-51 clone on top would be about 8.3 km/s. Even headed east, that’s nearly orbital velocity. So, not a regional prompt-strike weapon, not a testbed for an airbreathing first stage of a launch vehicle, not a general hypersonics experiment. Either an orbital spaceplane testbed, or a MaRV for a depressed-trajectory ICBM.

      It might sort of make sense of the observed signature, if we assume the failure occurred at stage separation – which is historically the case in ~1/4th of launch failures. Burnout for the first stage of a DF-4/LM-2C comes at ~2.3 km/s; depending on flight path that’s about right for an impact 600 km downrange. In this case, with a fully fuelled second stage still attached, so a bit off-target and making that impressive BFRC.

    • George William Herbert (History)

      Did some image moving and overlay; two of the debris.png photos were taken close together enough to roughly stitch up.

      I will try and sketch a map of the main site in a bit.

      Note that also, looking at debris.png, the lower-right quad image appears to be a different stage than the big one (upper right and left, lower left). The lower right stage has orthogrid aluminum, but its at 0-90 orientation (along the cylinder hoop / circumfrence, parallel to the axis) versus the +- 45 degrees of the other stage. Different structural layout.

      Could be different orthogrid in the upper tank in one stage, but my read is that it’s a different stage.

  4. RAJ47 (History)

    So you feel PRC has used LM-2F rather than LM-2C because its characteristics are far better?
    BTW, What engines does LM-2F stage1&2 use?
    What fuel does it use? Does it still use N2O4?
    How far would the LM-2F Stage 0 ( four boosters) and Stage 1 carry the Stage 2 & payload?
    Does YF-100 look similar to YF-21/YF-22?

    • Philippe (History)

      On the rocket itself, and sorry for having been unclear: I doubt we can get to that level of detail on the rocket type used (LM-2F vs LM-2C vs. something else), especially as we do not even have agreement on what stage crashed. I have presented a case for it being a second stage; other commentators here have raised serious objections to this theory in favor of a first-stage explanation.

      My comment on the power of the LM-2F was just to highlight that the LM-2F is a possible candidate rocket (more than powerful enough compared to the LM-2C and the Soviet/Russian system) that had lettering on the second stage.

      As a side note, the surplus rocket theory raised should also make us doubly careful with drawing inferences from the lettering pattern, since it may have been a “custom” paint job…

      On the other questions:
      1) This page on the LM-2F has lots of (unfortunately unsourced) details including on the engine, which could be a good place to start:

      And this page has a wonderful collection of Long March engine pictures, including one of the YF-100:

      The previously cited reference on the LM series confirms the fuel type for all LM series 1st and 2nd stages is N2O4/UDMH:
      It also gives launch capability to LEO as between 8,080kg and 8,600kg, depending on the subtype of LM-2F used. The LM-2F range for other mission types would of course heavily depend on the payload and trajectory.

  5. RAJ47 (History)

    Operational range of DF-5 is 12000km – 15000km.

  6. JO (History)

    Another point for 2 stage: the nozzle for that engine is a proportionately big bell nozzle typical of second stages.

  7. RAJ47 (History)

    The local picture of the launch in Taiyuan and others IMO are PSed.
    See the third set of pics at
    The second pic with Sun rays(in the third set) is obviously taken later as the plume has straightened out.
    But, when you look at the cloud formation of the two pics, the clouds are compact implying it is taken before.
    QED: The pics are PSed.