Why Is Mercury Liquid?

Or, Why Do Relativistic Effects Not Get into Chemistry Textbooks?

Lars J. Norrby
Royal Military College of Canada. Kingston, ON, Canada K7K 5LO

That mercury is liquid at ambient temperatures has been different. Au is an excellent conductor with a conductivity of
known since ancient times. The Greek name Hydrargyrum 426 kS m-'. Hg, on the other hand, is a much poorer conduc-
= "watery silver" (hence the symbol Hg) and the Latin tor with a conductivity of only 10.4 kS m-1. (All data given in
Argentum Viuum = "quick silver" show this as do the En- this article are taken from Greenwood and Earnshaw ( I )
glish and French names of the element alluding to Mercury, unless otherwise stated.)
the fast-footed messenger of the Latin gods. The Alchemists From a structural point of view we note that Cu, Ag, and
certainly knew mercury very well, especially its ability to Au have cubic and Zn and Cd (slightly distorted) hexagonal
dissolve gold, to amalgamate. As a matter of fact, amalgam- close-packed crystal structures. However, Hg is rhombohe-
ation of noble metals with subsequent thermal decomposi- drally distorted and the Hg-Hg distance in the less-than-
tion was a method of extracting such metals in use in the close-packed planes is about 16% "too large". Again, the
Mediterranean area already ahout 500 B.C. Probably all of metal-metal bonds in Hg are obviously weaker than they
us have, sometime, dropped a thermometer and tried to "should" be.
chase those evasive small droplets all over the floor. At room Although Au and Hg necessarily have very similar elec-
temperature there is no douht that mercury is liquid. But tron structures,
why? When I ask students, or colleagues for that matter, the
. .
answer goes "hm. .it is because. . .hm. .it has such a low
melting point!" No way!
Purpose we might expect that the slight difference, somehow, lies
I have consulted a fair number of currently used textbooks behind their strikingly different properties. How?
and "hibles" of inorganic chemistry including Greenwood
and Earnshaw (I) and Cotton and Wilkinson (2). Nowhere Anornalles
have I found an explanation of the well-known fact that Hg There are a number of unexpected periodic properties, at
is liquid with the exception of Mackay and Mackay (3),who least unexpected from a systematic point of view, when we
very briefly discuss this in a short section "Relativistic ef- look at the elements past the rare earths. Afamiliar example
fects". Cotton and Wilkinson do mention relativistic effects is the striking similarity between Hf and Zr. The lanthanoid
a few times, but they do not give any consistent account of contraction is the usual explanation for this, which is caused
the great influence of relativity on chemical properties. by the filling of the 4f orbital group (generally called a
However, there is an emharrassingly large literature includ- "subshell"). 4f electrons do not shield the nuclear charee
ing several excellent articles in this very Journal on this and nearly as well as do s and p electrons or even d electrons. one
related problems. See Pyykk6 (4, 5) and Suggested Read- alsospeaks of the lesser penetration of the4forhitals, which
ings. "Embarrassing" to us teachers of chemistry, that is. means that the 14 protons that are added as we go along the
How come this knowledge has not yet got into the main- rare earths are not fully shielded off hs the 14 4f electrons.
stream textbooks? This leads to gradually iarger effective nuclear charges and a
The purpose of this article is to present a fresh constella- corresponding contraction of the electron cloud. This is cer-
tion of experimental fads, theoretical calculations, and a tainly a true effect that is largely responsible for 71Lu3+
discussion of the chemical bonding in mercury that hopeful- being about 0.03 A smaller than 3gY3+,although there are 32
ly throws some new light on a number of classical issues in more electrons within the volume of the lutetium ion. The
inorganic chemistry. lanthanoid contraction is usually also held responsible for
the metallic radii of Ae and Au both beine- 1.44 A and those
Mercury and Gold of Cd and Hg both being 1.51 A.
It is most interesting to compare mercury with gold since Whv is Au eold-colored? Wbv is it not silver-colored?Why
the two elements are "next-door neighbors" in the periodic does AU have the highest electron affinity, -223 kJ molFi,
table but have dramatically different properties. The melt- outside the really electronegative elements? Higher than
ine ooints. for examnle. Au 1064 "C and He -39 "C. differ sulfur and almost as high as iodine. Why is TI stable in the
mire than for any dther pair of neighhoriG metalsin the oxidation state +I, P b in +II, and Bi in +III, while their
neriodic table (exce~tfor Li-Be where the difference also is congeners are more stable as +III, +IV and +V, respective-
about 1100 OC but fbr a different reason). The densities, Au ly?
19.32 and Hg 13.53 g cm-3, also differ more than anywhere The lanthanoid contraction alone does not explain all of
else. The enthalpiesof fusion are quitedifferent, Au 12.8 and these anomalies, even if it is a very useful concept. The
Hg 2.30 kJ mol-I. However, the entropies of fusion are very "inert 6s2 pair" introduced by Sidgwick in 1933 is another
similar, Au 9.29 and Hg 9.81 J K-' mol-', which demon- idea invoked; see, for example, an inorganic chemistry clas-
strates that here is actually "nothing wrong" with the ther- sic like Phillips and Williams (6). However, this latter con-
modynamic data of Hg. They consistently speak the same cept does not really explain why mercury is liquid or why
language: the bonding forces are much weaker in Hg than in Ph(I1) is more stable than Pb(1V). To find the real cause of
Au. These data just restate what we already know but do not all those anomalies we will have to look into an entirely
explain why mercury is liquid at ambient temperatures. different realm of science, namely relatiuity and its influ-
The electrical properties of gold and mercury are quite ence on chemical properties.

110 Journal of Chemical Education

Relatlvlty we usuallv see d.7.-orbitals oictured. This relativistic s ~ l i t t i-n e
of orbital energies is according to Dirac the real explanation
Einstein taught us with his special relativity theory of
1905 that the mass of any moving object increases with its
of the traditional snin-orbit couding- - enerw, which for a
heavy element like ~b amounts to as much a s 2 eV or nearly
speed, 200 kJ mol-I.
mml= m m s t / m Relatlvlsllc Effects
B o b calculated the speed of a 1s electron in the hydrogen In summary, following Dirac we may speak of three rela-
atom in its ground state as 11137 of the speed of light when it tivistic effects
is orbiting a t t h e Bohr radius 0.53 A. This speed is so low that (1) sl,n and p,/z orbitals contract quite a lot but pan to a much lesser
the relativistic mass is only 1.00003 times the rest mass. extent.
Although small, Sommerfeld took relativistic effects into (2) This in turn induces an extension outward of d and f orbitals
account when, in 1916, he refined Bohr's model introducing relative to s and p orbitals.
elliptic trajectories. However, when we turn to the heavy (3) "Spin-orbit coupling" ia actually the relativistic splittingof p, d,
elements 7 9 A ~80Hg,, and onward, the situation is quite dif- and f orbital energies. This effect hecomes large for the heavier
ferent. The expected average radial velocity for a 1selectron elements.
in an atom heavier than hydrogen is The sum of these effects becomes verv i m ~ o r t a nfor
t Au
(u,) '(Z/137)< and Hg, making the energy difference between the 5ds12 and
6s112 orbitals much smaller, see below. There is actually at
which
--- . for- He
-- - - means (801137bc = 0.58~.or 58% of the meed of least one more relativistic effect to consider called the Dar-
light. m,, &en becomes 1.23 m,.,. his in turn m e i s that win term. which accounts for the oscillators motion of the
the Bohr radius shrinks bv 23%. since the mass of the elec- electron ('Zitterbewegung") that becomesimportant near
tron enters in the denomihator: Thus the 1s orbitals in Au the atom nucleus. This term eives s electrons higher energy
and H e contract uerv much. Because all orbitals must be and expands s orbitals, which partially counteracts the ve-
orthogonal to one &other, an almost equally large mass- locity-mass contraction. It is implicitly included in "effect
velocitv contraction occurs for 2s,3s, 4s,5s, 6s, and 7s orbi- 2" ahow. The
- - - - ~Darwin term is usuallv exolicitlv invoked in
~

tals askell. Now, in order t o appreciate what really is going the "Pauli relativistic" treatment of the SchrBdinger equa-
on we need to look into Paul Dirac's relatiuistic quantum tion, which is less cumbersome to use than a pure Dirac
mechanics. model.
Dirac Quantum Mechanics Relatlvislb Calculations
Schrodinger quantum mechanics with its probability con- Relativistic enerw band structures for gold, see Takeda
tours, node patterns, and energy levels familiar to all under-
graduate students of chemistry is not adequate when treat-
(n, and other heavymetals and alloys have been calculated
bv a variety of methods, see PrykkB ( 4 , 5 ) and Christensen
ing the heavy elements. Within the framework of spin-orbit (8).~ l l u s t r ~ t i o of
n ssuch hand-stiucture calculations are not
coupling we learn that the angular and spin quantum num- easy to employ for the purpose of this article. They actually
bers 1 and s are "no good" for the heavy elements hut that need Brillouin zone theory and a whole host of concomitant
+
their vector sum j = 1 s still is, so that we get j-j coupling concepts t o be fully appreciated. Figure 1has heen chosen
instead of L-S (Russell-Saunders) coupling. instead as a simnler illustration of the main noints to be
~~~ ~~~~~~ ~ ~ ~

The electron spin was "invented" by the Dutch physicists made. It portrays the relativisticcalculationson the molecu-
Uhlenbeck and Goudsmit in 1925 to explain the fine struc- lar species AgH(g1 and AuHW by Pyykko (9) and Pyykkd
ture of the hydrogen spectrum. The Stern and Gerlach ex- and Desclaux (10). The energy differences hetween the 4d
neriment of
r-------~ ~- 1922.
-~ . where a beam of vaoorized silver atoms and 5s orbitals of Ae and .id and 6s of Au are obviously quite
was split in two by an applied external inhomogeneous mag- different, although their nonrelativistic counterpa& are
netic field. seemed to Drove this idea. The idea of electron
~~~ ~ ~

vew similar. There are differences, of course, between a free

spin, that we usually meet as part of the Pauli exclusion gold atom, oraAuH(g) molecule, and thecrystalline
principle, is a postulate added to the SchrBdinger solution solid, but the main features of the relativistic effects arestill
of the wave equation. I t is interesting to note that Pauli, who the same. For extensive discussions of this point see Chris-
was so instrumental in the development of the quantum tensen (8)and Koelling and MacDonald (11). Excellent arti-
mechanical model of the atom a n d a a s the one that intro- cles
-~~~on relativistic effects on eold chemistrv have been eiven
~~ ~ ~ ~ ~

duced the fourth quantum number in the early 1925, did not by Schwerdtfeger et al. (12, h). ~irst-prinkplecalcula'iions
himself believe in ~ h l e n b e c kand G~udsmit'sinter~retation bv Takeuchiet al. (141 demonstrate that the hieher cohesion
of it as an intrinsic motion of the electron. Finally, in 1928, energy of gold compared to silver is a relativistk effect.
Dirac made the synthesis between quantum mechanics and
relatiuity. He showed spin-orhit coupling to be a purely X-ray photoelectron spectroscopy (ESCA) experiments
relativistic effect and that electrons really do not spin a t all, on Au and other heavy metals and alloys thereof have shown
contrary t o what I am sure most chemists think. According that relativistic calculations are much closer to observations
to Dirac all electrons, including s electrons, have angular than are nonrelativistic ones. On the basis of both experi-
momentum, and there is no distinction between "orbital" ments and calculations one can conclude that the metal-
and "soin" aneular momentum. There is only one quantity metal bonds in Au(s) are brought about by the single 6s
labeled by theUangularmomentum quantum numb& j. ~ n s electrons with a 5d admixture but (almost) no 60.
orbital then gets the label sun I t is this angular momentum In the analysis of anomalous periodic proper&s a prob-
that operates in "electron spin" spectroscopic measure- lem still remains, for relativistic effects do not vary smoothly
ments. Furthermore, the Dirac treatment demonstrates that with 2. They rather seem to culminate for :&I. The relativ-
the p,, p,, and p, orbitals are quite different from our com- isticvelocity-mass contraction of the radiusof t h e & orbital
mon belief. They form two groups of orbitals (or "spinors"in of the free Au(g) atom has been calculated to about 16%(8).
the Dirac parlance) designated pl/2 and pa12 and labeled by Furthermore, the relativistic effects are overlaid with the
the angular quantum numbers j. Since s ~ and n pl/z atomic lanthanoid contraction. which in itself is relativistic to about
orbitals have the same angular dependence, a pl/z orbital is 158, as well as with an analogous 5d orbital contraction. For
in fact spherically symmetrical. I t is also lower in energy the 6s electron in solid gold the total stabilization of 2.8 ev
than the pa12 orbital, which is doughnut-shaped in the way (270 kJ mol-') arises 273 from relativistic effects and 113

Volume 68 Number 2 February 1991 111

 -eueldxa uommoa aq& 'a!ramouom L~?a!rls s! qarqm '(B)aH
p w ( 8 ) z ~uaamqaq aauarajj!p aq? w amleu ames aq? jo s!
'(B)Q aseqd wB aql u! a?Jawouow Klaqua l s o q e Bqaq Kq
s w a m Buome anb!un s! q q m 'Lmaram pue '(8)Znv 'awqd
w%aq? u! apaalom ram!p alqels papuoq L~Buoqse Bu!maoj
ploB uaaqaq aauaia~j!p aql ieql Kss ppoa auo 'Boleue
w merp OJ, 'sapads m!mpsa p w zaaI!s snofdopw aql u!
w q ? ralroqs saauqsrp Bu!puoq aql Bu!qsm ' + Z Z 8p~w Znv
qloq q w q w m se spuoq aq? ualzoqs qaajja a!lsv!$epr
lsql moqs (12) 'p la r a p a ! ~Kq suo!?epalua '7ae3 UI .spunod
-moa pgos p w uo!yIos snoanba qloq morj u m o q nam paap
-u! sa!aads alqqs Lraa 8 s! 'Znv q ? ! ~a!uoqaaIaos! s! qa!qm
'+,~BH 'raaamoH .Znv 0%a~qsmdmoaspuoq alBu!s luapaoa
Buoqs Lus mroj lou saop SH .(I)BH 0%pue aGqd ssB aqlo? aaS y q l JO 11e qi~miuamaase a~!ial!pnb u! s! 1 a m a j '8v
inolap e B u g q ramsw aql p q j Lsm aM iplnbq X ~ n a ~ as? w u! uuql p a r o q aiom K1[sqa4iaua are ieql spi!qro p!iqLq
XyM j a f q m s!q? $0 uogsanb [erzuaa aq? qnoqe moq 'MON k p 08 squeql saxa[dmoa q! u! Burpuoq $oa~eaoapaaunou
L P I ~ AmS~ew
~ I ~ SI &M
-old aiom S M O ~ Sn v ' a r o m r a q p a '(111)Bv ueql Kpuanbag
arom (1n)nv qnq (1)nv w q l (I)%v ralunoaua uago arom
am 'arojaraq& roj rardm~am pa!* p w puoaas aq? al!qm
.*
'Bv JO qeql wqq raBm1 rEpq%ls s! n v jo Braua uo!pz!uo!
p q aq& 'ssa~oloasmadds p w An u! sqrosqe dsB p w q
ra%q e B w e q 'raaamoq ' r a a p ~'pwq uo!lanpuoa aql JO
~
do$ a i l 78 I&[&IJ qaqm '[aaal !mrad aqi 01Brr~qiadsK[p!r$s
10 'laaal sg aql oq pal!xa are suortaala pg p w paqrosqs
s! 1qBg anlq uaqm s a s p n v 30 roloa aq& 'nv pus 8 v uaamlaq
saauarajj!p aq? jo pranas u!%qdxa oq alqe mou am a&
. ..
S ~ X B ~ I E U I W-I m
~ o d e ~ s l e dogz 3 r s w '. IOUJ r y s s z s z = ,w Z I z 12
8~3
= n ' B 1 '51 Un 3111089 U! ~ ~ 1 6 1 DLO!IBC3183 3 IS~A!IBIBIUOY PUB ?!IS!h!lBIB1
wo., (6lknv pue ( 6 1 ~ re 6 ~n q o w m.olaop aul l a sa16ls.e l m t q r ~r e.nB j
) 3 s 3 '7,em6!j
pueq alp Bu!moqsewads ( s ~ x v
we pl06 3!1181aw 10 se~n13n~ls
S'o
P'O
tion is that He&) does not exist because the third and the A), the energy needed to go from oxidation state +I to +I11 is
fourth valence electron would populate the antibonding lo* higher for TI+ than for In+. The sum of the second and third
orbital thus completely destabilizing a diatomic molecule. In ionizations energies is 4848 kJ mol-' for T1 and 4524 kJ
Hg the relativistically contracted 6s orbital is filled, and mol-I for In, i.e., a difference of about 7%.TI+ is isoelectronic
therefore the two 6s electrons do not contribute much to the with Hg, and the extra energy needed to remove two more
metal-metal bonds. This is obviously the opposite of gold. electrons to get to T P + obviously stems from the relativistic
Onemust conclude that the bondingis brought about largely contraction of the 6s orhital. The "inert 6s2 pair", SO often
by van der Waals forces, see Pyper et al. (221, and probably encountered but never explained in texthooks, is a relativis-
also through weak 6p orbital interaction. This is why the tic effect.
He-He bond is so weak.
-1nc~dentallY,this also explains why the electrical conduc- Bottom Llne
tivity is so much lower for mercury than for gold: the two 6s The influence of relativitv on the uronerties of h e a w ele-
electrons are rather localized and contribute only little to ments has been well underkood fo;ate1east 15 years.-~his
the conduction band. Mercury could in the vein of this anal- knowledge is naturally finding its way into research work on
ogy he called a pseudo noble gas. heavy metal chemistr;. See f(; example the works by Schro-
bileen et a1 (24-2fiL I t is high time that this is also reflected
Arnalgarnatlon in t h e teaching of chemistry a t the undergraduate leuel.
Why does liquid mercury dissolve gold and form amal- Take a look a t the literature suggested a t the end of this
gam? What is the chemical bonding like in an amaleam? article and enjoy the thought-provoking and elegant answers
%hy does Hg amalgamate well with Cu, Ag, Au, ana the t o the seemingly innocent question "Why is mercury li-
alkali metals? The reaction with sodium is well known from quid?" and other puzzles of the periodic table!
the chlor-alkali process, where an amalgam (with about 0.5%
Na bv mass) is formed at the liquid mercury cathode. Why Acknowledgmenl
does it react with the ammonit& radical N H ~ Why ? poor& I wish to thank Pekka PyykkB of the University of Helsin-
or not a t all with most transition metals? Let us try this for ki for manv valuable comments on this manuscriot. His
an answer. As mentioned in the introduction, it is easy to profound uiderstanding and elegant accounts of relativistic
retrieve gold from amalgam by simply heating it, so the effects in chemistry has been a great stimulation for me.
bonding forces between Hg and Au are not very strong.
Consider a hypothetical gaseous molecular species HgAu(g). -- -
In analow with He*+. which has been soectrosco~icallv Pitzer, K. S. Am. Chem. Ree. 1979.12,271.
PyykkB, P.: D ~ 1 e . u J.
~ .P. A m Chom. Res. 1979.12, 276.
Eharactezed, it wouidhave a three-electron bond, H W A ~ Mslli, G. L.,Ed. RelotiuisficEffarta in Atom. Moleeulas,and Solid*: Plenum: NewYork.
,om
with two electrons in a bonding 60 orbital and one electron in
Christensen, N. E. lnf. J . Quont. Chrm. 1984.25.223.
an antibonding 60* orbital. ~ L i would s be weaker than the Chrkfiansen,P.A.:Enoler, W.C,:Pitzer,K. S.Ann.Reu. Phys. Chem. 1985,36,407.
strong single bond in Au~(g)but stronger than the bonds Pyykka, P. Cham. Re". 1988.88.563,
between Hg atoms. Silver and gold provide one electron per
atom to the amalgam bonds. The same holds for the alkali Artlcles In This Journal
metals. Most transition metals react poorly with mercury Powel1.R. E. J. Chem.Edu. 1988.45.558.
Szeho. A. J . Chem Edue. 1989,46,678.
because they contribute twos electrons per atom, and we are MeKeluey,D. R. J.Chem.Edue. 1983,60,112.
back to t h e case of mercury itself; o d y very weak bonds Banna. M. S. J. Chem. Educ. 1985,62,197.
Mason, J. J . Chem. Educ. 1988,65,17.
would be possible. The ammonium radical also provides one
electron to the bond with Hg. I t seems that if the alloying Literature Cned
metal contributes one electron per atom, a good amalgam is I. Grenrund, N. N.;Eemahaw.A. ChemisfryollhsElrmonta: Pewamon: Oxford, 19%
formed
-~ but not if it contributes two. n119S

I t should be clearly stated that no unambiguous calcula-

1989:P 597.
tions have vet uroven the ideas discussed above. Detailed 3. Maeksy, K. M.: Maehy, R. A. Intmdvction LO Modornlnorganic Chemistry, 4th ed.;
relativistic hand calculations on solid mercury and amalgam Blsekie: Glaspaw and London, 1989:p 292.
are needed to substantiate these bonding ideas. T o carry out 4. Pyykko. P. Rolafivisfie Theory of A t o m ond Molecu1n.A Bibliogmphy 18161985;
Sorineer: Berlin. 1986. Contain8 more than 31W entries!
such calculations, i t is generally necessary to assume that the 5. ~yykkti,-P.Chom.Reu. 1988,88,563.
crystal structure of the metal is either cubic close-packed or 8. ~hillips,C. S. G.: willisms, R. J. P. Inorganic C h a m i s f ~ :Oxford University: Now
York. 1966;Vol. 2, Metah.
body-centered cubic. The rhombohedral distortion in crys- 7. Takeda,T. J.Phys.F.:MoiolPhy~. 1980,10,1135.
talline mercury is quite far from ccp, which therefore would 8. Christensen. N. E. Inf.J. Quclnf. Chom. 1984.25.233.
be a rather poor approximation. In the binary system Au-Hg
no intermetallic phase richer in mercury than Au2Hg has
heen confirmed. This stoichiometric comnound has a hexae- . ..
anal crystal structure and melts incon&ently a t 122 ' E . ~

Chem. Phys. 1989.9!'1762:

13. Schwerdfhger,P. J.Am. Chem. Soe. l989,111,7261.
The solid solubility of Au in Hg is negligible. See Rolfe and 14. Takeuchi, N.:Chan, C. T.: H o , K M.Phys.Rau.B. 1989,40,1565.
Hume-Rothery (23). Mercury-rich amalgams are therefore 15. Bagu%P.S.;Lee,Y. S.:PiUec,K.S, Chem.Phys.Let1. 1915.33.408.
more or less well-crystallized two-phase mixtures of AunHg 16. Chcistiansen, P. A ; Ermler. W. C.; Pitzer. K. S. Ann. Re". Phys. Chem. 1985.36.407.
17. Deadaux,J,P.: Pyykk~,P.Chem.Phy~.Leff. 1976.3W.
and Hg. Relativistic calculations on liquid mercury, where 18. Cole. L. A,; Perdew. J. P . Phm. Re". A 1982.25, 1265.
the structure of the liquid (a cluster of eight nearest neigh- 18, Werthelrn, 0 . K. In Solid State Chemistry Technique*; Cheetham. A. K.; Day, P..
Ed=: Clarendon: Oxford, 1987. Chanter 3.
bors a t 3.0 A) is t a k e n h t o account, would likewise be most 20. ~ h r i a t e n ~ ~ n , ~ . ~ . ; ~ o ~ l a r , ~ . ~ o ~1983,46,727.
.~toi~Comm.
useful. 21. Zieglor,T.;Snijders, J.G.;Baerends,E. J. J . Chem.Phys. 1381,74,1271.
22. -,.
Puoer.N.C.: - ~Grant.
~~~ . . I. P.:Gerbr.R.B.
23. Rolfe,'~.:~um-Rothe;.
Chrm.Phvs.Lefb. 1977.48.479.
W. J i e s s Comm ~et:1967.13, 1.
. .
The Inert 6s2 Palr 24. Giilespie, R. J.:Granger,P.,Morgan, K. R.:Schrobilgen, G. J.Inorg. Chem. 1984.23,
Finally, as a last example of relativistic effects on heavy nR7 .
25. Burns. R. C.; Oevereaur, L. A., Granger,P.: Schrahilgen, G. J. Inorg. Chem. 1985.24,
metal chemistry let us compare 81Tl and 431n. Although the 2615.
radius of the TI+ ion is larger (1.50 A) than that of In+ (1.40 26. Bj6rgvinsron. M.; Ssnyer,J. P.;Sehrohilgen. G. J. Inorg. Chem. 1987,26,741.

Volume 68 Number 2 Februaty 1991 113