Sie sind auf Seite 1von 11

Eur. J . Phys 10 ( I Y X Y ) ?

54-2W Prlnted In the UK

254

Early history of the physics and


chemistry of semiconductors-from
doubts to fact in a hundred yearst
Georg Busch
Laboratory for Solid State Physics, Swiss Federal Institute of Technology, 8093 Zurich, Switzerland

Received 17 March 1989

Abstract. Gray, DCsagullier and Volta discovered and Zusammenfassung. Gray, Dtsagullier und Volta
investigated electric conduction in solids. Davy found a entdecken und untersuchen als Erste die elektrische
decrease of the conductivity U in metals and Faraday Leitfahigkeit von festen Korpern. Davy beobachtet die
observed a strong increase with temperature in a number Abnahme der Leitfahigkeit U in Metallen und Faraday
of binary chemical compounds. Hittorf's measurements findet eine starke Zunahme mit steigender Temperatur in
on Ag2S and CuzS led to a linear relation of log U against einer Reihe von binaren chemischen Verbindungen.
1/T. The controversial case of Ag?S is described. Hall Hittorf's Messungen an Ag,S und CuzS fiihren zu einer
and Rowland discovered a transverse voltage of a current linearen Beziehung von log U als Funktion von l i T . Das
carrying metal film in a magnetic field. Riecke and Drude lange Zeit widerspriichliche Verhalten von AgzS wird
developed the first electron theory of metals and beschrieben. Hall und Rowland entdecken eine
Koenigsberger tried to explain the temperature transversale Spannung in einem stromdurchflossenen
dependence of the electrical conductivity by a Metallfilm in einem Magnetfeld. Riecke und Drude
dissociation theory. entwickeln die erste Elektronentheorie der Metalle und
Baedeker was the first to observe semiconducting Koenigsberger versucht, die Temperaturabhangigkeit
properties of CUI depending on the stoichiometric ihrer elektrischen Leitfahigkeit aufgrund einer
composition. Wagner proved that the conductivity of Dissoziations-Theorie zu erklaren.
Ag,S is essentially electronic and not ionic. Gudden Baedeker entdeckt als Erster die halbleitenden
suggested that semiconduction is the result of impurities Eigenschaften des CUI,die von der stochiometrischen
and imperfections in solids and Wagner and Schottky Zusammensetzung der Proben abhangen. Wagner zeigt.
developed their theories of lattice defects (Fehlordnungs- dass die elektrische Leitfahigkeit des Ag2S vorwiegend
Erscheinungen). Wilson presented the first band theory elektronisch und nicht elektrolytisch ist. Gudden vertritt
of intrinsic and extrinsic semiconductors. The existence die Auffassung, dass die halbleitenden Eigenschaften auf
of intrinsic conduction has been questioned by Verunreinigungen und Gitterbaufehlern der Festkorper
experimentalists and is verified only by the preparation beruhen, und Wagner und Schottky entwickeln ihre
and investigation of high-purity semiconducting elements. ersten Theorien der Gitterdefekte (Fehlordnungs-
Erscheinungen). Wilson entwickelt die erste
Bander-Theorie der Eigen- und Storhalbleiter. Die
Existenz der Eigenleitung wird von
Experimentalphysikern lange Zeit bezweifelt und erst
aufgrund der Herstellung und Untersuchung hochreiner
halbleitender Elemente bestatigt.

1. Introduction
'In a sketch of this kind, and in the time allotted for history of all the various insulated experiments that
it,Ishallnotbeexpectedtoproduceaminutehavebeenmade. My endeavour will be rather to fix
your attention upon leading and grand discoveries
BasedonanInvitedPaperdeliveredattheSpring which formtheepochsofthisscience.Manyhave
Meeting of theAmericanPhysicalSociety,NewOrleans, developed new objects in it, but a veryfew only
23 March 1988. principles.'
ascertainedhave
0113-0807/89/~W025J+ I 1 %U2 5 0 0 1989 1OP Publlshmg Ltd & The European Physlcal Soclety
history
Earlychemistry
and of the
physics of semiconductors 255

Obviously these are not my own words. This is tried in 1607 wereunsuccessful,buttheideawas
the introduction to a lecture by Humphry Davy [l] born.
beforetheRoyalSociety in London on ‘Electro- The first to introduce the conceptof an electrical
ChemicalScience’ in M O ! B u t in thefollowingI conductor was Jean
Theophile
Desagulliers.
shall keep Davy’s recipe in mind. Desagulliers was born in 1683 at La Rochelle. the
I began to workon semiconductors 50 years ago, hometown of thefamousSeignetteDynasty of
in March 1938 - on silicon carbide.This was a apothecaries, and he lived as a religious refugee in
dangerous enterprise.My friends warned me: work- London until 1744. In a ‘Dissertation sur I’electricite
ing on semiconductors means scientific suicide! As I des corps’,which is not equivalent to a PhD thesis in
was giving a seminar on semiconductors during the today’s sense of the word, he distinguishes between
war,acolleague of mine,professor of applied ‘electric’and‘non-electric’bodiesandintroduces
physics, got up and asked me: ‘What are semicon- theterm electric conductor (conducteurd’elec-
ductors good for? They are good for nothing. They tricite) explicitly for the first time. This work was
are erratic and not reproducible.’ Indeed this was awarded a prizeby the ‘Academie Royale des Belles
stillin aperiod of poorknowledge. Siliconwas Lettres et Arts’ at Bordeaux in 1742 [4].
believed metal;
be
to
a grey tin,
the
low- After searching the literatureof the 18th century
temperature diamond-type structure. of tin was said I came to the conclusion that Alessandro Volta was
to beasuperconductor,and in
silicon carbide the first to introduce the word semiconductor, or at
volume rectification was assumed. But the worst of least materials of semiconducting nature. This can be
all was a letter Pauli wrote to Rudolf Peierls from found in a paper read in English before the Royal
NewYork in1931; ‘UeberHalbleitersollteman Society in London on 14 March 1782 [ S ] .
nicht arbeiten, das ist eine Schweinerei, wer weiss, After the invention of his famous pile Volta had
ob es uberhaupt Halbleitergibt’ [2]. In loose transla- powerful sources of electricity at his disposal. Apart
tion this means: ‘On semiconductors one should not frommanyotherexperiments,heobservedthe
doanywork,that’samess,whoknowswhether speed with which an electrometer is discharged by
there are semiconductors at all’! touching its knobwith different materials. Metals do
it instantaneously, semiconductors slowly and insu-
lators not at all. To read Volta‘s long and numerous
2. Early research on electrical conduction papers is a difficult and time consuming task. He not
I think I have to begin with Stephen Gray [ 3 ] ,who only published in Italian,
but
also in French,
lived from 1666 or 1667 to 1736in Londonand German and Latin, and from translationsit is often
Canterbury. On8 February 1731 he wrote a letter to difficult to find out what the original version really
CromwellMortimer,thesecretary of theRoyal meant.
Society, containing the descriptionof several experi- The next person to be mentioned is Humphry
mentsconcerningelectricity.Hissource of elec- Davy again, the father of electrochemistry. H e had
tricity was a long tube of flintglass, closed at each built Volta piles up to several hundred volts, which
endwithacork.Whenheexcitedthetube by enabled him to do experiments under rather stable
rubbing, he observed that not only the glass tube, electrical conditions. H e wasinterested in the
but also the corks at the ends attracted light metal influence of the temperature on the conductivity of
foils or downfeathersverystrongly.Thisobser- metals, such as Cu, Ag, Sn, Pb, Fe and Pt. In 1840
vation led him to ask whether the electric charge he wrote: ‘The most remarkable general result that I
might be transmitted to other bodies also by other obtained by theseresearches,and which I shall
materials and at what distance. He began with short mention first, as it influences all the others, was, that
fir sticks and pieces of metals and ended up with a the conducting power of metallic bodies varied with
‘communication line’ of over 250 m. The ‘electrick temperature, and was lower, in some inverse ratio, as
effluvium’ was transported not onlyby a horizontal, thetemperature was higher [6].Ittooknearlya
but also by a vertical line of wood or packthread, century to explain these observations theoretically.
andalso in hoops of variousdiameters.Healso Michael Faraday [7] knew about Davy’s experi-
observed that bubbles of soapy water, i.e. a liquid, ments, and he extended them to a great number of
areabletocarryawayelectricity.Inadditionhe compounds that he had prepared himself. Not only
investigated whether the magnetic field of a piece of was he interested in the sulphides of various metals,
loadstone would influence the transmission of the butalso in oxides, carbonates, sulphates, nitrates,
electriccharge.Nodoubt,StephenGray is the halides,amongstothersandalso in periodide of
discoverer of electricconduction in solidsand mercuryHgI,.
Faraday soon realised that
the
liquids, without using this term in his letter, how- change of the conductivity with temperature of the
ever. In my opinion the intuition and originality of various compounds was opposed to whathis master
Gray’s experiments is comparable to Galilei’s ques- hadfound on metals. H e perceivedasamaterial
tion, whether light is a state of the space,or whether of particular interest
and
surprising
behaviour
it propagates in space with a measurable velocity. ‘sulphuret of silver’, i.e. Ag2S. Atroom temperature
Due to entirely inadequate means, his experiments its conductivity was very low, but at about 175 “C an
256 G Busch
abrupt rise to nearly 'metallic' magnitude occurred.
Unfortunately,Faradaypublishednoquantitative
resultsandnocurvestoshowthe resistivity asa
function of temperature.

3. The puzzle of Ag2S


Neither did Johann Wilhelm Hittorf, Professor of
Chemistry and Physics at the University of Munster
in Westfalen, Germany. His main work was dedi-
cated to the electric conduction of electrolytes, gases
andthephysics of cathoderays.In1851Hittorf
publishedapaperontheelectricconductivity of
Ag2SandCuzSasafunction of temperature [8].
Table 1 shows his results.The resistivity of his
specimens, which were prepared by direct reaction
of the elements, was compared with the length of a
platinum wire of equal resistance. Again, no plot is '1
i
shown. and so I was curious to do it myself, as I
preparedthispaper.Figure 1 showstheresult.
must confess I was flabbergasted as I saw the points
I
I A52S -

layingonastraightlineina log U against l i T


diagram for Cu2S. As I did the same forAg2S, I was
3 i 1 2
3
somewhat disappointed: No straight line. and what 1000/ T
about the jump at 170°C? Figure 1. Electrical resistivity of Cu,S and AgS as a
Ag2S is a very intricate substance indeed. It kept function of temperature plotted according table 1.
chemistsand physicistsbusy formorethan 120

years, and the literature aboutit is so enormous that


Table 1. Electrical resistivity of Cu2S and Ag,S as a I only can pick out the most important results. Ag2S
function of temperature (from Hiltorf [S]). occurs in nature as black, shiny crystals and under-
goes a first-order transition at170 "C from an orthor-
hombic p- to a cubic a-structure type. Above the
transitiontemperaturethesulphuratomsforma
body-centred cubic lattice and for the silver atoms
42 possible sites are available, which are occupied by
four atoms at random.
Hittorf believed that he had observedelectrolytic
conductivity. He could not know any better. But the
question,whethertheconductivity isreally ionic
orelectronicremainedcontroversialfornearlya
century.In 1921 Tubandt,SchibbeandEggert
[9,10] verified Faraday's law by an ingenious sand-
wich method, shown in figure 2 , based on the pure
ionic conductivity of silver iodide. Silver iodide acts
as a barrier for electrons and prevents the formation
of silver threads in the silver sulphide. By measuring
currentandtimeand by exactweighing of both
anode and
cathode,
Tubandt hoped to verify
Faraday's law exactly. However, in
1933 Carl

Figure 2. Tubandt's sandwich method to verify Faraday's


law in Ag,S [9. lo].

Ag A91 Ag 2s Ag
"?-a "?-a B-?-
Cothode i Ag' 1 Ag* e- I Anode
history
Early chemistry
of theand
physics of semiconductors 251

conduction.Themechanism of theelectricalcon-
ductivity of many other solids also remained obscure
for a long time, and the observation of the photo-
conductivity in selenium by Smith in 1873, [16], the
important discovery by Hall [l71 in 1879 and other
galvano- and thermomagnetic effects failed to eluci-
1 ’ e Sample 11 date the problem.

‘t ~

1
1
A Sample 15
A Sample 21
0 Sample 31
Sample 34 4. The Hall effect
“Klaiber 11929)
Edwin Herbert Hall was a student of Professor H A
e uel afterMiyatanil19551
Rowland at
the
Johns
Hopkins
Universityat
m
-0 Baltimore. Rowland is best known for his construc-
tion of high precision ruling machines and his reali-
-1 - sation of concave gratings, so important for spectro-
scopicwork.Whilereading Maxwell’s bookon
‘Electricityand
Magnetism’ in connection with
-2 - Professor Rowland’s lectures, Hall’s attention was
attracted by Maxwell’s statement, that ‘it must be
carefullyrememberedthatthemechanicalforce
-- 3 which urges a conductor carrying a current across
1 2 3 i the lines of magnetic force, acts, not on the electric
1000/T IK”1 current, but on the conductor which carries it . . .’.
Figure 3. Electrical conductivity of Ag,S as a function of
BothHallandRowlanddoubtedthetruth of
temperature (after [13]). Maxwell’s statement and Hall learned that Rowland
himself had made some hasty experiment to detect,
if possible, some action of a magnet on the current
itself, though without success. Hall asked Rowland
Wagner,[ll],Professor of Physical Chemistry in for permission to do a somewhat different experi-
Jena,repeatedtheexperimentsandcametothe ment: based on the following idea. If the currentin a
conclusionthatthevalidity of Faraday’slaw was conductor itself is attracted by a magnet. it should
only the result of a secondary reaction and that the be drawn to one side of a wire. and, therefore. its
ionic contribution in both the p and a phases was resistance should increase. Within the limits of error
less than a per centof the total current. This opinion no change of the resistivity was detected and Hall
wassupportedalreadyin 1929 by Klaiber[l21 in
Erlangen, who observed a Hall effect in both the a Figure 4. Hall coefficient of Ag2S as a function of
and p phases. His results were confirmed in 1959 by temperature (after (131).
Junod [13], and by Junod and co-workers [l41 who
showedthattheresultsareverysensitivetothe
preparation of the specimens, essentially purity and
deviationsfromstoichiometry.This is aserious
problem, because a considerable homogeneity range
exists in thephasediagram.Zone refining under
controlled sulphur pressure is necessary. It could be
shown that with the change of the structure at170 “C
a semiconductor-metal transitionoccurs.Junod’s
results of measurements of theelectricalconduc-
e Sample 11
tivity and the Hall coefficient as a function of tem- A Sample 15
perature are shown in figure 3 and figure 4. A Sample 21
The story of silver sulphide is not yet finished, 0 Sample 31
Sample 34
however. In the last twenty years several hundred . - Klaiber 119291
papers appeared on this particular substance treat-
ing the whole spectrumof its solid state physical and
chemical properties. A review article was published
by Rickert in 1967 [15].
To conclude, it appearstomethatHittorf in
1851 was the first to have measured quantitatively
the electrical conductivity of a semiconductor as a
function of temperature, albeit not knowing what
kind of chargecarrierswereresponsibleforthe
25 8 G Busch

came to the conclusion that it was more promising to


lookforatransversepotentialdifferenceat right
angles to the current and the magnetic field. This
was
the
arrangementformerly
conceived by
Rowland apparently, but tried on too thick a sheet
of metal. Following Rowland's advice Hall repeated
the experiment with a thin gold leaf mounted on a
plate of glass, and on 28 October 1879 the expected
effect was detected. The transverse potential differ-
ence was found to be proportional to the current and 0.02
to the magnetic field. In addition, Hall concluded -200 -100 0 100 200
thattheelectric field drivingthecurrent was de-
flected by anangleproportionaltothemagnetic
field,i.e.theHallanglewasconceivedalreadya
hundred years ago. Shortly after Hall published his
paper, Rowland repeated the experiment with iron
00051

0 0003
"-;.; .......
x. I
TI

and found oppositesigns of the potential differences - 200 -100 0 100 ZOO
observed for gold and germansilver [l8]. In retro-
spect, it would seem justified to name the observed
phenomenon the 'Hall-Rowland' effect, which is so
important to characterise the nature of electronic
conductors. (For a more elaborate account of the 0.00005 __/"

history of the Hall effect see [19]).


200 100 0 100 200
d eg
5. The electronic conduction of solids Figure 5. Electrical resistivity of Si, Ti and Zr as a
A decisive step forward in our recognition of the function of temperature (after [25]).
mechanism of the electric conduction in solids was
achieved by CarlViktorEmanuelRiecke's [20]
experiment at the Institute of Physics at Gottingen formula of the Hall coefficient, which is practically
in 1901. soonafter J JThomsondiscoveredthe identical to the one derived by Peierls in 1929 [24].
electron in 1897. Rieckeswitchedacylinder of However,neitherRieckenorDrude was ableto
copper between two cylindersof aluminium into the make any realistic predictions concerning the con-
main line of the battery, feeding the whole labora- centrations of the charge carriers as a function of
tory for a year, and he determined very carefully thetemperature.WhereasRieckeassumed a linear
weight of the copper cylinder before and after the increase of thenumber of mobilenegatively or
experiment. A total charge of 3.42 X lo6 C had flown positively charged particles
with
temperature,
through the copper cylinder, andif the current were Drudehesitatedtomakeanypredictions.Both
connectedwithatransport of copperatoms,a agreedaboutthedecrease of themobilities with
deposit of 1.14 kg of copper at oneof the connecting temperature to
explain
the
generally observed
electrodes ought to have shown up. But within the decrease of the conductivity of metals at high tem-
limits of errors of his measurements there was no peratures, but they were notin a position to give an
change of the weight. So he concluded that electrons explanation for theresistivity minimum of a number
mustcarrythecurrent.That this hypothesiswas of metals atlow temperatures, observed for instance
correct was shown by Tolman and Steward [2] only for titanium and zirconium found by Koenigsberger
in 1916. They were able to prove that the ratio eim and Schilling [2] in 1908. Their results are shown in
of the electrons in a metal was approximately identi- figure 5 .
cal to the ratio found for the free electron. Whether
by inertia experiments of this kind one determines
the rest mass of the electronor the effective mass was 6. Koenigsberger's dissociation theory
a puzzle for many years! IncontrasttoRieckeandDrude,Koenigsberger
Immediately after Thomson's discovery Riecke postulated that the mobile charge carriers were the
[22] in 1899 andDrude [23] attheUniversity of result of a dissociation of the atoms of a 'metallic'
Leipzig in 1900 proposed their first theories of the conductor into electrons and the remaining positive
electric conduction of metals by the assumption of ions according to the equation N = No exp( - Qii"),
an 'electron gas' carrying the current. It is remark- where Q is proportional to the dissociation energy.
able that Riecke already gave room for the existence JohannKoenigsberger,Professor of Physics at
of not only negatively, but also positively, charged the University of Freiburg im Breisgau was a versa-
carriers of possibly differentconcentrationsand tile scientist. H e was mainly interested in the elec-
unequalmobilities, which led him finally to a trical,opticalandthermalproperties of minerals,
Early history of the physics and chemistry of semiconductors 259

and he was an expert on the mineral world especi- zero, which means that all the atoms of a solid are
ally of the Swiss Alps. Together with a number of col- split into a positive ion anda free electron. For
laborators he studied mainly oxides and sulphides, ‘variable conductors’ Q was supposed to have a
such as Fe20s, Fe,O,, FeTiOs. FezTiOs, PbS, FeS, finite value, which means that the resistivity
MoS2, pyrrhotite and silicon. But apart from these decreasesexponentially with rising temperature.
solid-state problems, he was interested in spectro- With this assumption Koenigsberger indeed came
scopy, canal rays, thermalradiation, geophysical very close to thetruth; already in 1913! He also
phenomena and related subjects. His list of publica- observed that the values of Q and the mobility of the
tions is impressive not only with respect to its length, various variable conductors depended markedly on
but also to his originality. His dissociation theory their purity and the structural imperfections of the
to describe the temperature variation of solids de- specimens he investigated. Koenigsberger quite
serves attention. definitely belongs to the pioneers of semiconductor
F?r the temperature variation of the mobility of physics.
the electrons he proposed a decrease proportionalto But who introduced the term ‘semiconductor’ in
+
the inverse of an expression (1 arkpr’), which led todays sense of the word?
him to a formula for the resistivity In 1910 Weiss, a student of Koenigsberger, pub-
lished his doctoral thesis entitled‘Experimentelle
p=po(l+atk~t’)exp[Ql(t+273)]
Beitrage zur Elektronentheorie aus dem Gebiet der
which indeed leads to a minimum of the resistivity. Thermoelektrizitat’ [29]. He measured the thermo-
Koenigsberger andReichenheim [26] in 1908 electric power,theThomsonheatand the Peltier
claimed that the electrical resistivity of solids was coefficient of variouscombinations of different
represented quite generally by theirformula. metals with the oxides andsulphides of iron and
Although this conclusion certainly goes too far, it titanium and he compared his results with the elec-
supported the belief that the resistivity of metals tron theories of Riecke, Drude, JJ Thomson,
becomes infinitely large at zero absolute tempera- Lorentzand Koenigsberger. It is in this paper that
ture, an opinion which was shared even by theconcept of a‘Halbleiter’,i.e.a semiconductor.
Kamerlingh-Onnes,before he discovered supercon- appears for the first time,andthen again in a
ductivity in 1911 [27]! common publication by Koenigsberger and Weiss
Although Koenigsberger was unable to make in 1911 [30]. In a footnote of this paper,atten-
any prediction of the value of the dissociation tion is drawn tothe fact thatthe authorsdo not
energy Q , he classified the electronically conducting agree in every respect, but they do not specify their
solids into metals, insulators and ‘variable conduc- disagreement.
tors’, according to the value of Q [28]. For insulators In this context it is remarkable to notice that
Q tends to infinity, for metals at high temperature to Koenigsberger in a review article even in 1923 I311
included a chapter on ‘Variable Leiter’ (variable
JohannGeorg Koenigsberger 1874-1946 (by courtesy of conductors) and avoided the concept ‘Ha1b1eiter’
Mrs M J Loveday-Koenigsberger). (semiconductor)
throughout. So it appears that the
term ‘Halbleiter’ actually was coined by Weiss! But
it is obvious that variable conductors and semicon-
ductors have the same significance.

7. Baedeker’s research on CUI


An important step forward is due to Karl Baedeker
[31-351 who was Professor of Physics at the
University of Jena, Germany. Karl Baedeker was
born in 1877 as ason of Fritz Baedeker, the editorof
the world famous travel guides. He waskilled in
action in the first week of World War I at the age of
37. His list of publications is short, but his few
papers are fundamental. In 1911 he published a
book entitled ‘Die elektrischenErscheinungen in
metallischen Leitern’ [35] which served as a hand-
book in the field for more than two decades.
Baedeker was acquainted with Koenigsberger’s
observations on minerals and their erratic be-
haviour, whichis partly dueto inhomogeneity,
impurities,structural imperfections and unreliable
contacts. The poor reproducibility of the results is
260 G Busch
the main reason for the bad reputation semicon- that the conductivity is extremely sensitive to the
ductor research suffered for many years to come. iodine content of the specimens; an entirely new
Baedeker tried a new method. He produced thin phenomenon. By the samemethod he also produced
metallic layers on sheets of glass or mica by sputter- Ag2S. His results were compatible with the assump-
ing. Their thickness was estimated by weighing with tion of a transition from ionic to ‘metallic’ conducti-
a high sensitivity torsion balance. He then exposed vity at thetransition temperature which Faraday had
these films to oxygen or the vapours of sulphur, observed nearly 75 years earlier.
selenium, arsenic and iodine to produce the respec- The next step was the measurement of the Hall
tive compounds of copper, silver, cadmium, lead effect, which was considered as a proof of metallic
and thallium. Todaythisseems to be a trivial conduction. But the sign of the Hall voltage was
method, but in 1907 it was an ingenious idea. Most opposite to the one in bismuth, i.e. apparently due
of the layers weretransparentandcoloured,and to positively charged carriers. The Hall coefficient
their homogeneity was checked under the micro- again varied considerably with the iodine concentra-
scope. Burned-in platinum, prior to the sputtering, tion in agreement with the change of the conduc-
served as electrodes. tivity. Under the assumption of one kind of charge
The most important results were obtained with carrier, from the Hall coefficient Baedeker indi-
copper iodide. CUI is a colourless substance with a cates, one obtains an unrealistically low concentra-
surprisingly low electrical resistivity of approxima- tion of theorder of 10Xcm-’. According to our
tely 10”Qcm. Baedeker hesitated to ascribe this present knowledge this indicates two kinds of car-
low value to electrolytic conductivity, and in a sub- riers, electrons andholes. But Baedeker came to the
sequent paper he was able to show that the conduc- conclusion that CUI is a ‘metallic’ conductor with a
tion was of ‘metallic’ character, i.e. electronic. Left concentration of electrons increasing with tempera-
in open airat room temperature,theCUI layers ture. whichis essentially correct. (For amore
became practically insulators, but exposed to iodine detailed appreciation of Baedeker’s work see [37].
vapour or to an alcoholic solution of iodine the
conductivity immediately increased by several
orders of magnitude in a reversible way. This means 8. Gudden’s impurity hypothesis
During world War 1, research activities on semicon-
ductors practically came to astandstill,andwere
Kark Baedeker 1877-1914 (by courtesy of revived only in the early 1920s. In 1924Gudden [38]
Universitats-Archiv, Friedrich-Schiller-Universitat,Jena). at the University of Gottingen, a close cooperator of
R WPohl, published adetailed review on the
electrical conductivity of crystalline substances,
excluding ordinary metals. He gave a rather com-
plete description of the results obtained on ionic or
electrolytic conductors, but as far as semiconductors
are concerned, no real progress in understanding
their properties was made in the following years.
In 1930 Gudden published a new review article
on the electrical conduction of semiconductors [39].
In his opinion no chemically pure substance would
ever be a semiconductor. The observed properties
were believed to be due entirely to impurities, and
he came to the conclusion, that ‘semiconductors in
the scientific sense of the word - if they exist at all -
are by far scarcer than originally assumed’.

9. ‘Fehlordnungs-Erscheinungen’in solids
In theearly 1930s Frenkel [40],
Wagnerand
Schottky [41] and Jost (421 developed their models
of point defects in lattices, which led not only to an
understanding of diffusion and ionic conductivity,
but also to the explanation of the electronic conduc-
tion of ionic crystals. This was the birth of the so-
called ‘Fehlordnungs-Erscheinungen’ in solids,
which has become fundamental for the whole field
of solid state physics ever since. Figure 6 and 7 show
examples of ionic lattices with cationandanion
defects respectively. Defects in the anion lattice lead
history of the
Early chemistry
and
physics of semiconductors 26 1
U
-1
-2
-3
-4
-5
-6
-1
-8
Figure 6. Ionic lattice with anion vacancies: electron
(excess) conductor. l
1000 600400 200 100 20 0 -50 -100
SQO 300 I "C , 1 I
1 2 3 4 5 6
to electron conduction, defects in the cation lattice 1/rx103
to hole conduction. In opposition to Gudden's origi- Figure 8. Electrical conductivity of CuzO according to
nalidea,impuritiesarenotnecessary to produce different authors (after [%l).
semiconductivity therefore!
In the following years an enormous number of
mostly binary compounds - oxides, sulphides, sele- Defekt-Leitung'. i.e. excess and defect conduction,
nides, carbides, nitrides and others - were investi- the reason being missing positive or negative ions in
gated with respect to their electrical properties. It is thelattice.Examples of semiconductors of both
impossible to give a full account of all these results. naturesare givenin table 2 includingso-called
Veryoftentheconductivitiesfoundonthesame 'amphotheric conductors'. inwhich either positive
substance by different authors were in striking con- or negative ion sites may be unoccupied. Excess and
tradiction. Cu10is one of the most instructive exam- defect conductors show opposite signs of the Hall
ples, shown in figure 8. The conductivities of differ- coefficient. which is well known.
entsamplesmeasuredatthesametemperature This is in short the chemical aspect of the semi-
differed by 6 or 7 orders of magnitude! It soon was conductorproblem.
cannot
It be
emphasised
recognised that deviation from stoichiometry was the enough that semiconductor physics does not make
reason for these discrepancies. As shown in figure 9 sense on specimens of unknown purity and struc-
theconductivity of C u 2 0 increases with oxygen tural perfection. This remains true even today!
pressure,accordingtomeasurements by Diinwald
and Wagner [43].
Zincoxide is anoppositeexample.Measure- 10. Wilson's theory of semiconductors
ments by Fritsch [l41 in1935 on artificially grown Butwhatabout physics? ApartfromKoenigs-
crystals as a function of temperature (figure 10) and berger'sdissociationtheory,there was neithera
by von Baumbach and Wagner [43] show that the quantitative explanation of the exponential increase
conductivity rises rapidly with decreasing content of of the conductivity of semiconductors with tempera-
oxygen (figure 11). i.e. a deficit of theanion, in ture nor of the differentsigns of the Hall coefficient
contrast to Baedeker's measurement on CUI. Carl in different substances.
Wagner [46] in1933was the first to distinguish The breakthrough came withwave mechanics.
between * Elektronen-Ueberschuss-' and ' Elektronen- MaximilianStrutt [47], ayoungengineeratthe

Figure 9. Electrical conductivity of Cu,O as a function of


Figure 7. Ionic lattice with cation vacancies: hole (defect) oxygen pressure (after [43]).
conductor.

M' X- M'
x- l

l x- M+ GI n l -1 0
log po, ( m m Hgl
1 2
262 G Busch
1c
Table 2. Semiconducting compounds known in 1950.

t
'tl Excess conductors
TiO?. V Z 0 5 .Fe,O,. CuO. CulO,, ZnO, MOO,,
ScN. NblOc, CdO. CdS, CdSe. SnOl. SnSe, Cs2S, CszSe.

i
-1
BaO. BaTiO,, Ta,O,. WO,. Auz03.HgzS red, Hg,S
black. T1,O;. PbCrO,. Bi,Se,. UIO,, U 0 3
- -2 c
L

Defect conductors
-E -3 Cr,O,. MnO. COO. Co30,, NiO, CUI, Cu,O. Cu2S.
cD Cu&. Cu,Te, GeO, MOO?.Ag,O. SnO, SnS, Sb&.
TI@. T12S. Bi,Oi. Bi&, BilSe,. Bi2Te,
-," - L I
Amphoteric conductors
-5 L
Sic. Cr509. MnlO,, Mn,O,. Co,O,, G e , RuO?. Os,S1,
IrO?. PbO, PbS. PbSe, U 0 2
-6
t
-7 c
1
- 8 100 1
, , , 0, I-54
, -100
,, , o c ,
3 4 5 6 7 8 9 1 0 1 1 12
, , , -1po
dependence of the electrical conductivity of metals,
which lead to thewell known T" law at low tempera-
ture. But Bloch did not try to give any explanation
for the existence of metals, insulators or semicon-
1/~.103 ductors.
Figure 10. Electrical conductivity of Z n O as a function of This is themerit of AlanWilsonwho was a
temperature for various specimens. single crystals and student of R H Fowler at Cambridge. In 1931 he
evaporated layers (after [44)). spent about a year at the Leipzig Institute where he
foundaveryinspiring scientific atmosphere.He
Philips Laboratory in Holland and later a Professor drew the conclusions from Strutt's and Bloch's cal-
of Electrical Engineering at ETHin Zurich, was the culations and was the first to explain the difference
first to treat Schrodinger's equation for a periodic between metals and insulators based on his idea of
potential, which leads to
Mathieu's differential filled and empty energy bands. Alan Wilson, from
equation, solved by Floquet in 1873. Figure 12 gives London, is therealfather of the bandtheory of
the well known diagram, showing the typical finite solids which has dominated solid state physics ever
regions for real and imaginary solutions. At the end since. Bloch apparently was very reluctant at first to
of his short paper in 1928 Strutt modestly concluded accept Wilson's ideas, but he finally agreed.
that his results'mighthavesomebearingonthe In 1931 Wilsonproduced his classic papers
understanding of metallicconduction, or particu- [50,51] on semiconductors and
distinguished
larly superconductivity'! between'intrinsic'and'extrinsic'semiconductors,
Theproblemwastakenup by Felix Bloch by takingaccount of thepresence of donors and
148. 491 immediately. He was a post-doctoral fellow acceptors. The scheme is shown in figure 13.
at
the
Institute for Theoretical Physics atthe Wilson's model immediately explains the exponen-
tialincrease of theconcentration of thecharge
University of Leipzig,directed by Heisenberg. In
his famouspaper in 1928 heshowedthegeneral Figure 12. Floquet's solution of Mathieu's differential
mathematical character of the eigenfunctions of an equation. Shaded areas real. whit- areas imaginary
electron in a periodic potential: the Bloch functions. solutions (after [ 3 7 ] ) .
Especially. he developed a theory of the temperature-

Figure 11. Electrical conductivity of Z n O as a function of


oxygen pressure (after [45]).

10 14 1 8 2 62 2 30 -24 -16 -8 0 8 16 24 32 LO
log p c , ( m m ) A
Early history of the physics and chemistry of semiconductors 263
Physics Department of ETHZ for providing the literature
of past centuries.

References

[ 11 Davy H 1840 Historical Sketch of Electrical


Discoveries. Collected Works of Sir Humphry Davy
v01 111, ed J Davy, (London: Smith)
[2] Pauli W 1985 1930-1939 Scientific Correspondence
with Bohr, Einstein, Heisenberg and others v01 I1
(Berlin: Springer) p 94
131 Gray S 1731 Several Experiments concerning
Electricity Phil. Trans. R. Soc. 37 18-44
[4] DCsagulliers J T 1742 Dissertation sur I'klectricife des
corps (Bordeaux: Brun)
[ S ] Volta A 1782 Del modo di render sensibilissima la
piu debole electtricita sia naturale, sia artiliciale
Phil. Trans. R. Soc. 72 237-81
Alan H Wilson (by courtesy of Sir Alan Wilson). [ 6 ] Davy H 1821 Further researches on the magnetic
phenomena produced by electricity with some new
experiments on the properties of electrified bodies
in their relation to conductivity powers and
carriers,electronsand holes, with temperature. temperature Phil. Trans. R. Soc. 111 236-44
That the empty states in the nearly full band, the [7] Faraday M 1833 On a new law of electric
holes, are equivalent to positively charged carriers conduction Phil. Trans. R. Soc. 123 507-15
was shown by Heisenberg also in 1931 [52). The [8] Hittorf J W 1851 Ueber das elektrische
existence of electrons and holes explains the differ- Leitungsvermogen des Schwefelsilbers und des
ent signs of the Hall coefficient observed in semicon- Halbschwefelkupfers Ann. Phys., Lpz. 84 1-28
ductors. [9] Tubandt C and Schibbe G 1921 Ueber das
elektrische Leitvcrmogen des Schwefelsilbers und
As lateas 1939, justbefore World War 11, des Kupfersulfurs Z. Anorg. Chern. 117 1-47
Gudden gave aseminar in Zurich. He still was [IO] Tubandt C and Eggert S 1921 Ucber
extremely pessimistic with respect to semiconductor Elektrizitatsleitung in festen kristallisierten
physics, especially as far as theory was concerned, Verbindungen Z . Anorg. Chern. 117 48-58
and he did not believe in the existence of intrinsic [ 1 l ] Wagner C 1933 Ueber die Natur des elektrischen
semiconductors. He was right insofar as at that time Leitvermogens des a-Silbcrsulfids Z . Phys. Chern.
no clear cut case was known. The answer to this 21 42-7
question was given only by research on thesemicon- [ 121 Klaiber F 1929 Elektrisches und optisches Verhalten
ducting elements silicon andgermanium done at von Halbleitern I. Halleffekt und Leitfahigkeit des
Schwefelsilbers Ann. Phys.. Lpz. 3 229-52
Purdue University under the direction of Professor [ 131 Junod P 1959 Relations entre la structure cristalline
Lark-Horovitz [53],during and after the war. Since et les proprietes electroniques des combinaisons
this time semiconductor physics and chemistry have Ag2S. AgzSe et Cu,Se Helo. Phys. Acta 32 567-600
been among the most important fields in fundamen- [l41 Junod P, Hediger H , Kilchor B and Wullschleger J
tal and applied science of condensed matter. 1977 Metal-non-metal transition in silver-
chalcogenides Phil. Mag. 36 941-58
[IS] Rickert H 1967 Elektrische Eigenschaften von festen
Acknowledgment Stoffen mit gemischter Elektronen- und
I am indebted to Professor Heinrich Medicus, Rensselear Ionenleitung Festkorperprobleme 6 85-105
Polytechnique Institute, for suggesting me as a speaker at [ 161 Smith W 1873 Effect of light on selenium during the
the New Orleans meeting, to Professor Si Yuan and Dr passage of light Nature 7 303
Fritz Hulliger of the Laboratory of Solid State Physics and [ 171 Hall E H 1879 On a new action of the magnet on
Professor Charles Enz. University of Geneva, for reading electric currents Phil. Mag. 9 225-30
the manuscript and to Dr Ulrich Helg,librarian of the [ 181 Rowland H A 1880 Preliminary notes on Mr. Hall's
recent discovery Phil. Mag. 9 432-4
[l91 Buchwald J 1979 Centaurus 23 1
[20j Riecke V E 1901 1st die metallische Leitung
Figure 13. A H Wilson's first energy band diagram of an verbunden mit einem Transport von Metallionen?
extrinsic semiconductor. Phys. Z. 2 639
1211 Tolman R C and Stewart T D 1916 The
electromotoric force produced by the acceleration of
metals Phys. Reo. 8 97-1 16; 1917 The mass of the
electric carriers in copper, silver and aluminum
Phys. Reo. 9 164-8
264 G Busch
[22] Riecke V E 1898 Zur Theorie des Galvanismus und [36] Baedeker K 1911 Die elektrischen Erscheinungen in
der Warme A n n . Phys. Chem. 66 353-89, 545-81; metallischen Leitern (Braunschweig: Vieweg).
1906 Ueber die Elektronentheorie des Galvanismus [37] Kaiser W 1978 Karl Badekers Beitrag zur
und der Warme Jahrb. Radioakt. Elektron. 3 24-47 Halbleiterforschung Centaurus 22 187-200
[23] Drude P 1900 Zur Elektronentheorie der Metalle [38] Gudden B 1924 Elektrizitatsleitung in
Ann. Phys., Lpz. 1 566-613 Kristallisierten Stoffen unter Ausschluss der Metalle
[24] Peierls R 1929 Zur Theorie der Ergeb. Exakten Naturwiss. 3 116-59
Galvanomagnetischen Effekte 2. Phys. 53 255-66 [39] Gudden B 1930 Ueber die Elektrizitatsleitung in
[25] Koenigsberger J and Schilling K 1908 Ueber die Halbleitern Sitzungsber. Phys.-Med. Soz. Erlangen
Leitfahigkeit einiger fester Substanzen Phys. 2. 9 62 289-302
347-52 [40]Frenkel J 1926 Ueber die Warmebewegung in festen
[26] Koenigsberger J and Reichenheim 0 1906 Ueber und fliissigen Korpern Z. Phys. 35 652-69
ein Temperaturgesetz der elektrischen Leigfahigkeit [41] Wagner C and Schottky W 1931 Theorie der
fester einheitlicher Substanzen und einige geordneten Mischphasen Z . Phys. Chem. 11 163-
Folgerungen daraus Phys. 2. 7 570-8 210
[27] Kamerlingh-Onnes 1904 The importance of accurate (421 Jost W 1933 Diffusion and electrolytic conduction in
measurements at very low temperatures Commun. crystals (ionic semiconductors) J . Chem. Phys. 1
Phys. Lab. Univ. Leiden Suppl. 9 1-30 466-75
(281 Koenigsberger J 1914 Das elektrische Verhalten der [43] Dunwald H and Wagner C 1933 Untersuchungen
variablen Leiter und deren Beziehung zur uber Fehlordnungserscheinungen in Kupferoxydul
Elektronentheorie Jahrb. Radioukt. Elektron. 11 und deren Eintluss auf die elektrischen
84- 142 Eigenschaften 2. Phys. Chem. B 22 212-25
[29] Weiss J 1910 Experimentelle Beitrage zur [44] Fritsch 0 1935 Elektrisches und optisches Verhalten
Elektronentheorie aus dem Gebiet der von Halbleitern. Elektrische Messungen an
Thermoelektrizitat Inaugural Dissertation Zinkoxyd Ann. Phys., Lpz. 22 375-401
Albert-Ludwigs Universitat. Freiburg im Breisgau 1451 von Baumbach H H and Wagner C 1933 Die
(Emmendingen: Dolter) elektrische Leitfahigkeit von Zinkoxyd und
[30] Koenigsberger J and Weiss J 1911 Ueber die Cadmiumoxyd 2. Phys. Chem. B 22 199-211
thermoelektrischen Effekte (Thermokrafte. [46] Wagner C 1933 Theorie der geordneten
Thomsonwarme) und die Warmeleitung in einigen Mischphasen 2. Phys. Chem. B 22 181-94
Elementen und uber die experimentelle Prdfung der [47] Strutt M J 0 1928 Zur Wellenmechanik des
Elektronentheorien Ann. Phys., Lpz. 35 1-46 Atomgitters Ann. Phys., Lpz. 86 319-24
[31] Koenigsberger J 1923 Metallische Leitung [48] Bloch F 1928 Ueber die Quantenmechanik der
Handbuch d Elektr. und d. Magnetismus Bd I11 ed Elektronen in Kristallgittern 2. Phys. 52 555-60
L Graetz (Leipzig: Barth) pp 597-724 [49] Bloch F 1930 Zum elektrischen Widerstandsgesetz
[32] Baedeker K 1907 Ueber die elektrische Leitfahigkeit bei tiefen Temperaturen Z. Phys. 59 208-14
und die thermoelektrische Kraft einiger [50] Wilson A H 1931 The Theory of Electronic
Schwermetall-Verbindungen Ann. Phys., Lpz. 22 Semiconductors I Proc. R. Soc. A 133 458-91
749-66 [51] Wilson A H 1931 The Theory of Electronic
[33] Baedeker K and Pauli E 1908 Das elektrische Semiconductors I1 Proc. R . Soc. A 134 277-87
Leitvermogen \on festem Kupferjodur Phys. Z . 9 (521 Heisenberg W 1931 Zum Paulischen Ausschliess-
43 1 ungsprinzip Ann. Phys., Lpz. 10 888-904
[33] Baedeker K 1908 Eigentumliche Form elektrischen [53] Lark-Horovitz K 1954 The present state of physics
Leitvermogens Phys. 2. 9 431-5 The New Electronics ed F S Brackett (Washington.
(351 Baedeker K 1912 Kunstliche metallische Leiter DC: American Association for the Advancement of
Phys. 2. 13 1080-2 Science) pp 57-157