4. 2003-003B's A/m Compared with that of Shuttle's TPS
2003 Mar 12: added information to Summary on materials undergoing radar signature testing, per CAIB press conference of 2003 Mar 11.
2003 Mar 23: added analysis and discussion of the possibility that the object was a Carrier Panel. (Introduction and Section 4.) Revised Section 1 to reflect recent reports confirming my assumption that the object separated at a low velocity relative Columbia, i.e. came loose.
The purpose of this hobbyist exercise was to see what might be learned by analysing the public information available in February 2003. I plan to revise it as additional facts warrant. This is not an effort to compete with CAIB's experts, whose training, experience and access to information and analytical tools far exceed mine.
Based upon the available facts, and what I believe are reasonable assumptions (Section 1), I have deduced 2003-003B's approximate initial orbital elements (Section 2), and from them, its area to mass ratio (A/m) (Section 3). An object's A/m determines the amount of drag that it experiences. The greater the A/m, the greater the drag, and the faster the rate of decay from orbit.
2003-003B's A/m, about 0.040 m2/kg, was 10 to 20 times greater than that of a payload or rocket body, indicating that it was debris-like. That explains its rapid decay from orbit, just three days after it separated from Columbia.
In Section 4, I evaluate the possibility of 2003-003B having been an element of Columbia's TPS (Thermal Protection System) - tiles, blankets and carrier panels. This is difficult to do with confidence because the attitude of 2003-003B is unknown. Its attitude would have determined its effective area subject to drag. A tile orientated face into the direction of orbital motion would experience far greater drag than one oriented edge-on.
2003-003B's rotation would have averaged out its effective area, but there is still a large uncertainty due to the lack of information on the orientation of its axis of rotation. The CAIB may eventually determine the orientation of the axis; for now, the best I could do was to evaluate A/m over the range of possible orientations.
Initially, I found 2003-003B's A/m to have been most consistent with the physical properties of the shuttle's RCC (Reinforced Carbon-Carbon) panels and its High Density HRSI (High Temperature Reusable Surface Insulation), but only in the range between average and maximum values of effective area.
In the CAIB press briefing of 2003 Mar 11, Maj. Gen. Barry listed carrier panels among the items to undergo radar signature tests. They are removable access panels that carry tiles or thermal blanket. In the CAIB press briefing of 2003 Mar 18, board member James Hallock speculated that 2003-003B could have been a carrier panel, as reported by MSNBC.
This prompted me to consider the aluminium carrier panels. I found that without any tiles, they could have closely matched the A/m of 2003-003B. Carrier panels that retained a 1/8" or 1/4" densified layer of High Density HRSI tile, could also have matched. Carrier panels that retained 3" thick High Density HRSI tile would not have been a good match.
Under certain circumstances, FRCI (Fibrous Refractory Composite Insulation Tiles) and Low Density HRSI could also have matched 2003-003B's A/m - mainly if in sheets of several tiles.
AFRSI (Advanced Flexible Reusable Surface Insulation Blankets) were not likely to have been consistent with 2003-003B.
It is interesting that 2003-003B appears to match the characteristics of the densest TPC components: RCC, High Density HRSI and Carrier Panel, but it could be something else with a similar shape and mass. Refinement of 2003-003B's A/m and determination of its attitude could put it outside the characteristics of some or all of these components.
In a discussion of my analysis on sci.space.shuttle, an issue was made of the ability of non-electrically conducting TPS to reflect radar. At least one person argued that the TPS would not reflect radar; at least one argued that it would reflect radar. The aforementioned MSNBC article quotes CAIB board member James Hallock reporting the preliminary results: "We’ve performed tests in a radar chamber. A tile doesn’t register, a blanket doesn’t even register." Clearly, the aluminum carrier panels would have registered. No word as yet on the RCC panels.
Hopefully, the CAIB's experts will have sufficient aerodynamic and radar data to determine with high confidence the parts of Columbia that could account for the mystery object, and those that could not.
A check of the satellite catalogue reveals that no previous shuttle mission that did not conduct an EVA or satellite deployment resulted in the cataloguing of unaccounted debris. Most of the catalogued debris were items confirmed lost by astronauts during EVAs (STS 51-I, STS 88, STS 102). A few pieces were shed as a result of an anomalous payload deployment (STS 51).
Two pieces of debris were catalogued subsequent to the EVA of STS 106; however, I have been unable to find any reports from NASA linking them to the EVA. Analysis of their orbital elements places one of them in the vicinity of the shuttle during the EVA. I plan to study this further.
2003-003B may well be unique; however, it was only found after it had decayed, as a result of an unprecedented post-flight search of archived radar observation logs, motivated by the loss of Columbia and her crew.
Would a similar retrospective of previous shuttle missions have turned up other unaccounted debris? What would have been their physical properties? If the archival data exists, it might be worth the effort to find out just how unique 2003-003B may have been.
My analysis is based upon the following facts and assumptions:
The following facts are based on the CAIB press briefing of 2003 Feb 25:
- first radar detection on 2003 Jan 17
- tracked by Eglin AFB in Florida, Beale AFB in California, Cape Cod AFS in Massachusetts, and the Navy Fence [NAVSPASUR]
- decayed on 2003 Jan 20 over the South Pacific Ocean
- initially semi-stable, in a slow rotation
- appears to be a piece about 0.3 meters by 0.4 meters of undetermined composition
The following facts are based on the CBS News report of 2003 Feb 08:
- separating from Columbia at about 5 m/s ... about 24 hours after launch
The following facts are based on the CNN News report of 2003 Feb 10:
- "A radar installation at Eglin Air Force Base in the Florida panhandle -- linked to U.S. Space Command headquarters in Cheyenne Mountain, Colorado -- recorded an object slowly moving away from the shuttle."
The following facts are based on the MSNBC report of 2003 Mar 18:
- not a propulsive separation, i.e. came loose and moved away from Columbia solely as a result of their different rates of orbital decay
- separated from Columbia some time before first detection
- first radar detection at 15:44 UTC pass of Eglin AFB (25 h after launch)
- separating from Columbia at about 5 m/s at time of first detection
- decayed during approximately the middle rev of those that passed over the South Pacific
- to simplify calculation, assumed that separation occurred at an ascending node
By trial and error, I found that separation at the ascending node of 09:31 UTC fits the relevant facts and assumptions, resulting in the following 2-line orbital elements:
STS 107 1 70000U 03017.39685426 .00068256 00000-0 14399-3 0 00 2 70000 39.0160 224.2509 0012417 356.5886 3.4837 15.97601887 05 2003-003B 1 70001U 03017.39685427 .02200000 00000-0 46411-2 0 08 2 70001 39.0160 224.2509 0012417 356.5886 3.4837 15.97601887 06
I chose to have the separation coincide with an ascending node for computational convenience, which is reasonable given the low precision of the available information.
STS 107's elements have been propagated from the nearest earlier epoch to the assumed moment of separation. At that moment, 2003-003B would have had the same elements, except for the much greater rate of decay shown.
By trial and error, using Alan Pickup's SATEVO orbit decay propagator, I found that ndot/2 of 0.022 rev/d2 fit my assumption of decay on Jan 20, during one of the middle passes over the South Pacific.
By the time of first detection, about 6 h after separation, 2003-003B would have been about 3 km below and 60 km (or 7.8 s time) ahead of Columbia, and separating at about 5 m/s, solely as a result of their different rates of orbital decay.
It is worth noting that if I had assumed that separation occurred close to the time of the first observation, then the rate of decay would have been about 10 percent greater.
The uncertainty of the time of decay (i.e. the location over the S. Pacific) results in +/-10 percent uncertainty in my estimate of the rate of decay.
For a circular orbit, area to mass ratio is given by the following formula:
A/m = 5.0237*10-9 * ndot/2 / ( Cd * rho * n(4/3) )
where: | A/m | = area / mass, m2/kg |
ndot/2 | = one half rate of decay, rev/d2 | |
Cd | = drag coefficient, assumed = 2.2 | |
rho | = atmospheric density, kg/m3 | |
n | = mean motion, rev/d |
I estimated the atmospheric density experienced by 2003-003B using the National Space Science Data Center's web-based MSIS-E-90 Atmosphere Model. The period covered by the model ended 2002 July, so I ran it against six historical proxy dates with 10.7 cm solar flux1 and geomagnetic Ap index similar to those of 2003 Jan 17-20 UTC.
Those dates were 1972 Jan 21, 1978 Jan 3, 1983 Jan 19, 1993 Jan 14, 1994 Jan 6, 1999 Jan 14. I chose January dates to ensure that solar illumination was similar to that experienced by 2003-003B.
For all six proxy dates, I computed the mean atmospheric density during 2003-003B's first revolution after the epoch of its orbital elements (Section 2). The mean values ranged between 2.5*10-11 kg/m3 and 3.7*10-11 kg/m3, with an overall mean of 3.1*10-11 kg/m3, used to compute A/m:
A/m | = 5.0237*10-9 * 0.022 / ( 2.2 * 3.1*10-11 * 15.976(4/3) ) |
= 0.040 m2/kg |
The object's A/m was 10 to 20 times that of typical payloads or rocket bodies, which combined with its small size, suggests that it was debris-like.
To accurately estimate the A/m of the TPS requires knowledge of their effective area for drag, which in turn requires knowledge of the attitude of 2003-003B. All that has been reported by the CAIB is that it rotated slowly. It is reasonable to assume that it rotated about a transverse axis. The CAIB may eventually determine the orientation of the axis of rotation, but for now, the best that can be done is to evaluate A/m over the range of possible orientations.
The minimum effective area occurs when the axis of rotation is parallel to the direction of orbital motion. It is equal to length x thickness.
The maximum effective area occurs when the axis of rotation is perpendicular to the direction of orbital motion. It is equal to (length + thickness) * width * 2 / PI.
In this section, A/m has been computed for the minimum effective area, the maximum effective area, and the arithmetic average of the two.
For physical data on the Shuttle TPS (Thermal Protection System) I relied mainly on NASA's News Reference Manual.
They are removable access panels that carry tiles or thermal blanket. The ones in question are located between the wing leading-edge RCC panels and the rest of the wing. Their approximate dimensions are 0.00318 m x 0.102 m x 0.559 m (1/8" x 4" x 22"), and they are made of aluminium, of density about 2,700 kg/m3.
Attached to the carrier panels on the under-side of the wings are High Density HRSI (High Temperature Reusable Surface Insulation Tiles), of thickness approximately 0.0762 m (3 inches). and density about 352 kg/m3.
The following table shows the effective A/m of a carrier panel with areal dimensions 0.102 m x 0.559 m, with and without the 0.0762 m HD HRSI tiles:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | ||||
Carrier - m | HRSI - m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.00318 | 0.00000 | 0.487 | 0.0018 | 0.0036 | 0.0363 | 0.0747 | 0.0191 | 0.0392 |
0.00318 | 0.07620 | 2.010 | 0.0444 | 0.0221 | 0.0413 | 0.0205 | 0.0428 | 0.0213 |
In the sci.space.shuttle discussion cited earlier, it was suggested that if a carrier panel became separated from the shuttle, only the densified layer of the tiles would be likely to remain attached. Below is the A/m that would result from 0.00318 m (1/8") or 0.00635 m (1/4") densified layers, of density assumed to be 50 percent greater than the bulk density of a whole tile:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | ||||
Carrier - m | HRSI - m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.00318 | 0.00318 | 0.582 | 0.0035 | 0.0061 | 0.0366 | 0.0628 | 0.0201 | 0.0345 |
0.00318 | 0.00635 | 0.677 | 0.0053 | 0.0079 | 0.0368 | 0.0543 | 0.0201 | 0.0311 |
RCC (reinforced carbon-carbon) panels have a density of about 1600 kg/m3, and range in thickness from 0.00635 m to 0.0127 m.
The following table shows the effective A/m spanning the known range of thickness of a flat fragment of RCC (curved fragments are also quite possible), for the reported areal dimensions of 2003-003B, 0.3 m x 0.4 m:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.00635 | 1.219 | 0.0025 | 0.0021 | 0.0776 | 0.0637 | 0.0401 | 0.0329 |
0.01020 | 1.958 | 0.0041 | 0.0021 | 0.0783 | 0.0400 | 0.0412 | 0.0210 |
0.01270 | 2.438 | 0.0051 | 0.0021 | 0.0788 | 0.0323 | 0.0419 | 0.0172 |
The following table shows the effective A/m spanning the known range of thickness of HRSI, for the areal dimensions of a single 0.15 m x 0.15 m tile:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.02540 | 0.201 | 0.0038 | 0.0189 | 0.0167 | 0.0833 | 0.0103 | 0.0511 |
0.03690 | 0.292 | 0.0055 | 0.0189 | 0.0178 | 0.0611 | 0.0117 | 0.0400 |
0.06470 | 0.512 | 0.0097 | 0.0189 | 0.0205 | 0.0400 | 0.0151 | 0.0295 |
0.12700 | 1.006 | 0.0191 | 0.0189 | 0.0265 | 0.0263 | 0.0228 | 0.0226 |
It would take several HRSI tiles to account for the area of 2003-003B. I do not know whether or not several could detach together as a sheet, but for the sake of argument, the following table shows the effective A/m spanning the known range of thicknesses of HRSI, for a sheet having the reported areal dimensions of 2003-003B, 0.3 m x 0.4 m:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.02540 | 1.073 | 0.0102 | 0.0095 | 0.0812 | 0.0757 | 0.0457 | 0.0426 |
0.02740 | 1.157 | 0.0110 | 0.0095 | 0.0816 | 0.0705 | 0.0463 | 0.0400 |
0.05100 | 2.154 | 0.0204 | 0.0095 | 0.0861 | 0.0400 | 0.0533 | 0.0247 |
0.12700 | 5.364 | 0.0508 | 0.0095 | 0.1006 | 0.0188 | 0.0757 | 0.0141 |
The following table shows the effective A/m spanning the known range of thicknesses of FRCI, for the areal dimensions of a single 0.15 m x 0.15 m tile:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.02540 | 0.110 | 0.0038 | 0.0347 | 0.0167 | 0.1526 | 0.0103 | 0.0937 |
0.12700 | 0.549 | 0.0191 | 0.0347 | 0.0265 | 0.0482 | 0.0228 | 0.0415 |
It would take several FRCI tiles to account for the area of 2003-003B. I do not know whether or not several could detach together as a sheet, but for the sake of argument, the following table shows the effective A/m spanning the known range of thicknesses of FRCI, for a sheet having the reported areal dimensions of 2003-003B, 0.3 m x 0.4 m:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.02540 | 0.585 | 0.0102 | 0.0174 | 0.0812 | 0.1388 | 0.0457 | 0.0781 |
0.06100 | 1.405 | 0.0244 | 0.0174 | 0.0880 | 0.0626 | 0.0562 | 0.0400 |
0.10450 | 2.408 | 0.0418 | 0.0174 | 0.0964 | 0.0400 | 0.0691 | 0.0287 |
0.12700 | 2.926 | 0.0508 | 0.0174 | 0.1006 | 0.0344 | 0.0757 | 0.0259 |
The following table shows the effective A/m spanning the known range of thicknesses of HRSI, for the areal dimensions of a single 0.15 m x 0.15 m tile:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.02540 | 0.082 | 0.0038 | 0.0463 | 0.0167 | 0.2035 | 0.0103 | 0.1249 |
0.12700 | 0.411 | 0.0191 | 0.0463 | 0.0265 | 0.0643 | 0.0228 | 0.0553 |
It would take several HRSI tiles to account for the area of 2003-003B. I do not know whether or not several could detach together as a sheet, but for the sake of argument, the following table shows the effective A/m spanning the known range of thicknesses of HRSI, for a sheet having the reported areal dimensions of 2003-003B, 0.3 m x 0.4 m:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.02540 | 0.439 | 0.0102 | 0.0231 | 0.0812 | 0.1851 | 0.0457 | 0.1041 |
0.09650 | 1.668 | 0.0386 | 0.0231 | 0.0948 | 0.0569 | 0.0667 | 0.0400 |
0.12700 | 2.195 | 0.0508 | 0.0231 | 0.1006 | 0.0459 | 0.0757 | 0.0345 |
The following table shows the effective A/m spanning the known range of thicknesses of AFRSI, for the reported areal dimensions of 2003-003B, 0.3 m x 0.4 m:
Thickness | Mass | Min Effective Area | Max Effective Area | Ave Effective Area | |||
m | kg | m2 | m2/kg | m2 | m2/kg | m2 | m2/kg |
0.01140 | 0.197 | 0.0046 | 0.0231 | 0.0786 | 0.3989 | 0.0416 | 0.2110 |
0.02410 | 0.416 | 0.0096 | 0.0231 | 0.0810 | 0.1945 | 0.0453 | 0.1088 |