What makes it so devious is a real gold bar is purchased with the serial numbers and papers, then it is hollowed out, the gold is sold, the tungsten is put in, then the bar is closed up. That is a sophisticated operation.
No need for there to be "duplicate serial numbers". Just take a cast of the "real gold bar" with its original serial numbers, melt it, put the tungsten bars in the mold, and refill the mold. Casting can capture VERY fine detail. The gold volume replaced by the tungsten stays in the crucible. You pour that into a different mold, and sell that separately in a black market where no serial numbers are needed. But given the crudity of the pictured example and the very thin layer of gold over the tungsten, a PROPERLY SELECTED XRF unit would have no problem picking this geometry up. Judging from the photo, there is probably no more than a mm or two between the tungsten and the gold surface.
And as "PapaBear3625" points out, there are other physical properties that work. I suspect that ultrasound would also show things up. I'm sure gold and tungsten have drastically different sound transmission characteristics.
Yep. Tungsten velocity of sound = 5174 meters/second. Gold velocity of sound = 3240 meters/second. Ultrasound should show tungsten cores quite readily.
Probably much quicker to put the bar in a CNC machine and hollow it out, than it is to go through the molding and casting process. Then polishing up the cast bar carefully so as not to remove details or leave marks.
A CNC machine can probably hollow out the original bar in seconds.
I doubt these reputable houses are accepting these bars without doing their normal checks.
I think the pictures and vid clearly show that the gold was not poured around the tungsten bar. The tungsten bar is just inside the gold bar. Gaps can be seen where the fit is not perfect.