The Achilles' Heel in this process is, of course, that it is limited by my ability to deliver a clear and understandable answer. Let me try to do better this time.
Is this statement accurate?
According NIST:
Summary of metallographic analysis – Core Columns
• Two core columns in impact area with sufficient paint
• Columns 603 (floors 92-93) and 605 (floors 98-99)
Source:
NIST
Not all of them were so heavily damaged that they could not be identified.
Besides if they were so heavily damaged, then through process of elimination by using the rest of the steel that was identifiable you can determine the location of the steel. I don't blame NIST for this however, it wasn't them who ordered the removal of the debris without testing and detailed examination.
Indeed it wasn't NIST who ordered this, nor did it even happen. But your summary is not quite complete. There are
four core fragments, positively identified, from the floors of interest, the two above from WTC 1 and also two others from WTC 2, both pieces of the same column. Those were from column 801 basically at the point of impact.
The two column fragments you list above were two floors above, and three floors below, respectively, the initiating event. Like I said,
"no surviving core fragments from within a floor of the initiating event" above. With respect to the WTC 2 fragments, that column was actually destroyed at impact, so it's impossible for any explosive to have detonated on it later.
The point is that we simply don't have the steel we would want to test in the first place. That moots any testing for chemical residue. The only way you're going to find chemical evidence of explosives is if either (a) you test
all of the unidentified steel and hope to get lucky, or (b) propose an explosives scenario that affected a very large area of the Towers. Case (b) is pretty silly, but I'll continue to work on (a) through the rest of the problems.
Most if not all scientific experiments have the potential for contamination. This does not mean experiments should not be conducted. This excuse can now be rejected. I would refer you to the chemical tests done in the case of TWA 800 as example.
TWA 800 had over 95% of its debris recovered, unlike our case, and even so there are many conspiracy minded folks who reject the test results, and still insist it was shot down. You are correct that nearly all scientific experiments can be contaminated, so whether or not it's worth running the test depends on an estimation of the
degree. That depends on experiment design. An experiment that is only 50% likely to be correct is useless.
Not only that,
increasing the sample size will help to determine the presence or absence of explosives or to reduce measurement error (e.g., perhaps through increased sample size or longer observation time) so that the densities are more tightly defined around the mean measurement as detailed
here.
Yes, I'm quite familiar with statistics and propagation of errors.
Increasing the sample size in this case is impossible, because we are already considering all of the steel pieces of interest. If we follow the Case (a) situation above, we can increase the "sample size" only by considering more steel, most of which we know was not involved with any hypothetical controlled demolition. We therefore increase the size of the sample but also decrease the signal strength.
A good rule of thumb for population statistics is that our uncertainty scales roughly as one divided by the square root of the number of samples. Unfortunately, by considering the whole volume of steel, of which only a fixed percentage at most is really from the region of interest, we decrease the signal linearly with inclusion of new but unaffected samples. This results in a net loss of accuracy for this particular case. Again, this is not the case if one expects the
entire Tower was wired with explosives, but again, that would be silly.
This goes back to your erroneous first point.
The probability of observing a false positive is referred to as the “false positive probability” or “false positive rate” and is equivalent to the specificity of the detector
not the condition of the steel. Source:Here.
I wasn't clear enough on this point. What I mean is the following: Explosive charges would have been designed to damage the steel columns and/or connections. After suffering collapse, there would be additional damage to the steel columns, possibly breaking off affected areas or mixing them with other debris. The condition of the steel surely cannot
help our detection effort, and it seems likely to hinder it.
Chemical detector specificity will depend on many factors, but concentration of chemical species is sure to be dominant. Abrasion, fracture, and impact would all contribute to reducing the chemical concentration.
The location of the piece tested has no bearing on whether explosive devices were used or not. I'm kind of surprised you listed this as a reason for the inconclusive result.
Again, let me try to be more clear. Supposing you believed you had evidence of explosive materials, your next step would be to
construct a hypothesis to see if it made sense. If your somewhat unreliable signal happened to also have geometrical correlation -- all from the same floor, all from connections, a "box" pattern, etc. -- that would further support the explosives theory, whereas a random or badly skewed distribution of signals would not.
Because the pieces we are testing, by definition, have lost their identifying marks, we cannot attempt this step. We therefore require an unusually clear result through chemical testing alone. Unfortunately, we won't get one.
That of course depends on the temperature of the steel and the properties of steel and whether or not it the sample was even exposed to fire.
I'm referring primarily to the fires
after collapse, which affected the overwhelming majority of the structural steel. Fires before collapse presumably would not have affected any explosive residue, since it hadn't exploded yet, although the steel we are most interested in all did burn before collapse which certainly wouldn't help. It's also more a factor of the properties of the chemical signature than the steel; the chemical signature is all but guaranteed to be more susceptible to fire than the steel itself.
The fires of course were not on all floors below the impact area. This of course can be rejected as well as the above point.
Again, the fires
after collapse, which burned for weeks, affected a vast amount of the steel. You must have overlooked those fires. In addition, the chemical plume from the fires would have affected
all of the steel, even those not directly hit by the flames, so our chemical confounding signal is expected to be ubiquitous.
I think we might all agree that it was unfortunate the amount of steel that was sampled represented a quarter to half a percent of the 200,000 tons of structural steel used in the construction of the two towers. The rational scientific approach is the one found in the process of forensic examination via chemical testing.
You've confused "sampled" with "recovered." The amount of steel recovered, i.e. retained for future study, massed about one half of one percent. But the amount checked ("sampled") was much higher, perhaps 80%. It's not realistic to retain all of the steel, seeing as how this would be roughly equivalent to two aircraft carriers worth of steel. Hence a downsampling was applied up front, perfectly ordinary.
The team looked through all the steel at the end of the salvage process, only retaining those items that could be identified, for two purposes. For estimation of as-built physical properties, they retained relatively undamaged samples. For determination of failure modes, they retained any piece that showed an unusual failure mode including excessive burns, unusual fracture, examples of connection failure / weld failure / necking / buckling and what have you.
Explosives would have created unique failure modes detectable by visual inspection. Not a single example was found. We may assume that the people inspecting were not infallible and might have missed some of them, so to be pessimistic, we may assume that only, say, 25% of these items would have been found. Still, we found
zero.
The bad logic that follows is if we don't see any evidence on this tiny percentage, then it didn't happen. So what we are left with is 99.95% of the steel unexamined visually and 100% of the steel that was not exposed to tests that could identify the chemical residue of explosives.
Again, no. Try about 20% worth of steel not examined visually.
So let's work through a prospective experiment design to see how this would work. Suppose we have a chemical test that is 95% accurate in sensing explosives residue. Let's say we have a number of samples that we believe may all have been exposed to explosives. Up front, we might say that, if 15% or more of the samples test positive, then we will proclaim that there were explosives. This would correspond roughly to "three sigma," e.g. three times the detector uncertainty, so our confidence in the experiment would be approximately 97% at minimum. That's a pretty good experiment.
However, what we have is different. Suppose we can't identify the steel uniquely. We have two overlapping populations, one that
may have been exposed, and another that
was not exposed -- but we cannot tell them apart. Let's be charitable and say that, in our situation, the interfering population is nine times the size of the desired population.
Now, even before running the tests, we expect that at most only 10% will test positive -- this is as good as the signal is going to get. We can no longer use a "three sigma" criterion, because we expect it to fail. To make this work, we have to relax the test, to perhaps 7.5%, a minimum confidence level of only about 70%. This result is on the ragged edge of being conclusive. And this is how a weak signal leads to an increase in
both false-negative and false-positive results.
And this assumes that all the steel actually exposed to explosives carries the signal. Suppose only half of the steel retains the signal for whatever reason -- volatile chemicals, lucky shielding, galvanization, handling issues, whatever. In this case, it is
impossible to design an experiment with a confidence level greater than sheer chance. You may as well not even run it.
In our situation, the number of steel samples we have positively identified is
zero within the area of initiation, and only nine from the total impact area, four core samples and five perimeter samples. If we relax identification and consider all the steel, the entire volume of steel is about one or two floors worth out of 110, for no better than two percent. Impact damage, fires before and after collapse, collapse itself, contamination due to fires, and contamination due to salvage operations and handling could have degraded the signal by a factor of ten or more though it's hard to guess without identifying the specific test.
In other words, our situation is basically hopeless.
Here's a personal anecdote: I was once held up at an airport security check-in by a chemical detector. Was I carrying explosives? Nope. The sensor had been tricked by a book I'd gotten from my grandfather, and apparently its pages had absorbed traces of nitroglycerin from his blood pressure medication... If you tune your experiment to be super-sensitive in response to a weak signal, such as we have with the WTC steel, you will get false alarms all over the place. It's not a problem in airline security because you accept a high false alarm rate, and you can repeat the experiment with more definitive methods. In this case, we cannot repeat the experiment, and we have no more definitive tests. It's a waste of time.
So a metallurgist would have to test the metal to determine if it were steel or not to corroborate what was witnessed? Or a metallurgist would have to be present to examine the metal before you accept it as molten steel?
Now, this situation is one where we
do accept a high false alarm rate and can follow up. We accept that eyewitnesses looking at hot, flowing stuff is pretty imprecise. But that steel would have cooled...
...so where is it?
Again this goes back to the illogical statement: if it isn't seen, then it didn't happen.
What exactly was found? Are there any descrpitions during the debris removal early on or later on fitting the items the descriptions below match?
I haven't heard of any such findings. Remember, all of the debris was sorted, down to individual bits of victims. They should have found cooled blobs of steel, but they didn't. You can't prove a negative, but you can put an upper bound on its probability.
If not, are all the accounts below lies? Even the first hand accounts?
One of them is. Mark Loizeaux later confirmed that Chris Bollyn had misquoted him. The others are hardly definite -- in particular, that last one; if the steel had melted, how could they pull it out of the ground?? Logically, there's no reason to suspect they might not be mistaken.
So the following comment from R. Mackey regarding testing of explosives in the WTC attacks.
1. He chemically test in the impact area only. Thereby ignoring a vast percentage of the tower itself. Tsk, tsk. Not a proper forensic investigation in my opinion when compared to other investigations that used the chemical testing for explosive residue.
Again, you only test where you think there were explosives. That would be the initiating area only. Otherwise, you're deliberately polluting your signal. Even if you think there
might be explosives elsewhere, you should design your experiment to test where there
must be explosives.
2. By doing this, he can claim the tests would be inconclusive because of the heat's effect on the chemical properties of steel despite the laboratory evidence not displaying high temperature exposures of the confiscated debris.
Again, heat would have affected all of the steel, not just the point of initiation. The fires after collapse are the major problem.
3. But just in case tests are positive, they are inconclusive because we aren't sure the part of the tower the steel came from. This is nonsense of course and the point has no bearing on the tests themselves nor the scenario itself.
You're reasoning backwards. In order to determine the tests are positive, you would want to corroborate using whatever means are at your disposal. Geometry would be one of them, and this is impossible.
4. Despite the sample size fallacy mentioned by Ryan, an increased sample size would reduce the false positive probability of the detector and serve to remove a very good criticism of the process, the extremely small percentage size of the steel samples that was investigate.
Again, this is only true if
all of the steel you test could bear traces of explosives. It would not. All of it was, however, affected by the false signal sent up by burning plastics. You'd increase your susceptibility to experimental error while actually reducing your signal to noise.
5. Ryan is satisfied that there were no explosive devices because there was no visual indication in any of the. 05% sample size from the recovered steel.
I would argue that a .05% of a sample is too little of a sample to conduct a proper investigation in the first place. But that is neither here nor there at this point.
I'd agree, except it isn't 0.05%, or even 0.5% which is what I think you meant to say, but actually around 80%. That's a pretty respectable sample size. And even if this wasn't the case, you still are left with the fact that confirming evidence does not exist. At best what you have is doubt, not evidence.
Finally, there is the avoidance to offer empirical proof for either the molten steel or the presence of explosives: the chemical tests that would prove one way or another.
As explained, I don't how this test can remotely be expected to succeed.
I hope that explanation was clear. Future discussion should probably be split to its own thread. I'd also be glad to help you with your persistent formatting problems, but that's probably a matter for PM.
unless there are more here than I'm aware of) in this forum would have to speak towards that. And at any rate, there's a good argument in saying that it's best to keep the failure from happening in the first place, which again swings focus back towards the impact zone. And that leads us back to what NIST's goal was in studying the failure. It wasn't to prove to anyone anywhere that there were no explosives involved. It wasn't to legitimize one side of an internet phenomenon over another. It was to understand how the impacts and the fires caused the structure to fail so that they gain a better understanding of such failures and can improve construction standards to make such failures unlikely. That's the intent of the study. So given that, why would it be incumbent on NIST to study steel outside the impact zone? If the general cause of failure outside the impact zone is understood - again, floors beneath that zone were overloaded - then why bother inducing the expense and logictical problems of keeping the steel, let alone study or run tests on it?
