Why isn't the silicon analogue of benzene flat?

Why isn't the silicon analogue of benzene flat? As far as I understand, silicon and carbon make similar structures and it doesn't make sense that its benzene analogue should be substantially different.

Quoted from the article:

Chemists in the UK have constructed a structural analogue of benzene made from silicon atoms. The molecule is not flat like benzene, but it reveals a new type of aromatic stabilisation.

This 'new' type of aromatic stabilisation is said to occur only around the seat of the chair structure.

I am thinking that silicon's bonds are too weak to be kept in a double bond, and therefore it doesn't have enough energy to change the shape of the bonding and is thus quite unstable.

• Well, for starters, I wouldn’t consider the two silicon atoms outside of the plane to take part in aromaticity, because they have four bonds each. So the actual aromatic part would be only four silicons and flat. – Jan Jun 9 '15 at 17:34
• @Jan Yep. Hence my comment on the seat of the chair structure. – Ali Caglayan Jun 9 '15 at 17:37
• It doesn't look like new type - 6 electrons in flat ring... True hexasilabenzene would be unstable as Si-Si pi bonds are weaker than sigma. – Mithoron Jun 9 '15 at 18:10
• strongly related chemistry.stackexchange.com/questions/28702/… – Mithoron Jun 9 '15 at 18:12
• At some level, an equivalent question is why silicon does not form a graphite-like phase. It really likes those tetrahedral bonds and will accept no substitutes! Note that the 'chair' structure is present in the diamond cubic structure. – Jon Custer Jun 9 '15 at 23:14

The linked article is a short summary of a Science article: "A Tricyclic Aromatic Isomer of Hexasilabenzene" by Kai Abersfelder, Andrew J. P. White, Henry S. Rzepa, and David Scheschkewitz; Science 2010: 327 pp. 564-566.

They propose a new term, "dismutational aromaticity":

The highly dispersed $\ce{^29Si}$ nuclear magnetic resonance shifts in solution ranging from +125 to −90 parts per million indicate an inhomogeneous electron distribution due to the dismutation of formal oxidation numbers as compared with that of benzene. Theoretical analysis reveals nonetheless the cyclic delocalization of six mobile electrons of the $\pi$-, $\sigma$- and non-bonding type across the central four-membered ring.

They point out that while $\ce{Si=Si}$ bonds were ignored for a long time, multiple compounds have been prepared, even including delocalization, like 1,2-disilabenzenes.

They prepared $\ce{Si6R6}$ with "Tip" or 2,4,6-triisopropylphenyl groups.

The surprisingly stable 3a can be exposed to air for hours as a solid or for minutes in solution without detectable changes.

Long story short, there is delocalization around the central 4-center ring:

Note that there is a 1,2-shift occurring, and that 2 $\ce{Si}$ atoms have two substituents, 2 have one substituent, and 2 have no substituents.

The paper argues that a fully planar hexasilabenzene might be possible, but:

the choice of precursor may be important in any synthetic approach.

Presumably the choice of substituent is also relevant to prevent the rearrangement here.