P A P E R S
Vacuum-depssited metal film resistors
by G. S D D U L and B. A. PROBYN, Research Laboratories, Edwards High vacuum Limited, Crawley, Sussex
MS. received 10th July 1961
Abstract
Although wcuum-deposited thin metal films have been gold films on cleaved rocksalt and on silver surfaces. Preston,
studied for many years, it is only recently that their Gillham and Williams (1955) have deposited gold films of
anomalous electrical properties have been turned to high electrical conductivity by using a glass substrate pre-
practical account. In this work the most important viously coated with a thin layer of bismuth oxide. Such f&-,s
requirements for the deposition of metal film resistors are have a resistivity which closely approaches the bulk value,
discussed, and an annealing cycle up to 300-350" c in air but considerable crystal imperfections, e.g. dislocations and
is proposed to stabilize the resistance of nickel-chromium stacking faults, are Still Present. It is thus unlikely that the
alloy films, preventing further large irreversible change boundary effect contributes to the high resistivity of very
within the temperature range of normal use. The thin metal films as significantly as earlier workers had
stability of annealed unprotected firms within the range thought.
1C-400 Qlsquare was determined over a period of 2000 h Although the resistivity of films is higher than that of the
on load (1 w/in') and 8000 h at no load. Frequency &a- bulk metal, the use of thin films for resistance elements is
grams show the number of resistors measured, and their limited by a f d a m e n t a l problem. Metal films having a
percentane resistance change. Resistors at no load useful resistance value are usually unstable. This effect has
I been studied by Vand (1943) who has shown that the irre-
versible decreases of resistance, which occur upon heating a
thin film, can be explained by re-ordering of the crystal
lattice due to a 'decay of the lattice defects'. These changes
hcreasez by an average of 1 % compared with a very
small mean decrease in the same resistors on load. The
temperature coefficient of resistance of films in the range
100-150Rlsquare is shown to vary between 0.002 and
0.1 % deg-'c and is related to the rate of film deposition.
Starting with an alloy of 80120 nickel-chromium, the
lowest values of temperature coeficient of resistance are
obtained for rates of deposition lower than 5 As-' and this
effect is discussed in terms of the oxidation of the film
and the separation of the alloy during evaporation.
1. Introduction
T the beginning of this century, Thomson (1901)
proposed a theory to explain why the resistivity of A thin metal films should be greater than that of the
bulk metal. He deduced that the mean path of the charge
carriers (later to be called the free electrons) between suc-
cessive interactions or collisions with the atoms in a thin
conductor was shortened by the closeness of the boundaries.
The resistivity was increased because of the energy lost at
each collision. This idea was extended by Planck (1914),
Fuchs (1938), Lovell and Appleyard (1937) and others, who
tried to relate the increase of resistance to the decrease of
thickness. Unfortunately many of these attempts were based
upon the assumption of an ideal thin film with parallel
boundaries, free from defects and regular in structure. In
view of present-day knowledge available from numerous
electron microscope studies of film structure, it is not sur-
prising that the above work was unable to explain the very
wide variation of experimental results. It is now known that
&ns deposited by evaporation in a vacuum are usually
highly imperfect, and this fact is sufficient reason for the
resistivity to become much higher than the value of bulk
resistivity for the deposited metal.
Recent experimental work has shown that under carefully
controlled conditions, ordered films can be deposited on
surfaces of regular crystal structure. For example, Bassett
and Pashley (1961) have succeeded in growing oriented
I
are much larger than resistance changes due to the normal
(positive) temperature coefficient of resistance of the material.
Vand has proposed that a new metal film has lattice disorder
with vacant sites and excess atoms situated so near that the
unstable equilibrium is easily disturbed. Separate energy
levels exist, corresponding to different types of defect in the
structure, and a definite temperature is required to remove
the defect at each energy level. The resistance of the film
falls to an irreversible value, and this structure should be
stable at temperatures below its previous annealing point.
This stability is not realized in practice because of oxidation
or absorption of atmospheric water vapour, but stability can
be improved by suitable protection.
Vand's work was based upon the annealing of films
deposited at liquid air temperatures. When films are
deposited at about 20" c or higher, the decreases of resistance
observed during subsequent heating are less marked, and
tend to be compensated by resistance changes due to further
oxidation, agglomeration or recrystallization.
After annealing, the resistivity of a thin metal film remains
higher than the bulk value, but the temperature coefficient
of resistance is usually lower. It has been shown by Holland
and Siddall (1953) that the increased resistivity of gold mS
within the range 40-400 A can be related to the decrease Of
temperature coefficient of resistance. The contributions to
the resistivity from lattice imperfections, boundary effect and
gas inclusion are greater in a thin film and, over a limited
temperature range (after annealing), are reasonably
pendent of temperature. Thus the temperature-dependent
part of the resistivity is comaaratively less than in the bulk
material.
It has also been observed that the temperature coefficient
of resistance of a very thin conductor is often negative. For
example, thin nickel and iron ~s deposited by catho&C
sputtering were shown by Itterbeeck et al. (1951) to Possess a
high negative temperature coefficient of resistance. Holland
BRITISH JOURNAL OF APPLIED PHYSICS 668 VOL. 12, DECEMBER 1961

VACUUM-DEPOSITED M E T A L FILM RESISTORS
and Siddall (1953) have pointed out that such films were
deposited under conditions in which oxidation was likely to
occur; thus the observed properties could be attributed to the
formation of a partially oxidized metal layer behaving as a
semiconductor.
The erratic behaviour of thin metal films is well known
and is the subject of an extensive literature, but as shown
by the foregoing, a better understanding of film properties
is beginning. Summarizing the Introduction, it can be stated
that the anomalous electrical properties of films are prin-
cipally due to their structural imperfections and to the
thermodynamic instability produced when metal vapour is
abruptly condensed to the solid phase. The changes of
resistivity and temperature coefficient of resistance, which
occur upon heating or ageing a film, arise from the re-ordering
of the structure, the relief of high internal stresses and the
further oxidation or gas absorption of the film. These
changes are parallel to those occurring in fme resistance
wires upon annealing after cold-working. However, the gas
absorption and higher degree of lattice imperfection in
vacuum-deposited films cause much greater variation of
properties. There is an increasing amount of evidence that
high stability resistance films can be obtained by correct
annealing treatment and suitable protection. It is the purpose
of this paper to describe a method of making reasonably
stable resistance elements by the vacuum deposition of
nickel-chromium alloy on glass and to discuss their properties
in terms of the processing conditions.
2. Practical requirements
Substrate
The surface of the supporting substrate should be smooth
and uniform, and both chemically and mechanically stable
at temperatures up to about 350"c in atmosphere and
vacuum. Any variation of surface smoothness gives a corre-
sponding variation of film resistance value, because the film
is thin enough to be greatly affected by the state of the
surface. For example, a film of resistance as low as
10 Q/square on a polished glass surface may be discontinuous
when deposited under identical conditions on a fmely ground
glass surface.
It is a characteristic of a film deposited from the vapour
that the grains tend to grow on surface prominences, which
trap the atoms first arriving there and act as centres for
nucleation. Films as thick as 1000 .i may be discontinuous
when deposited on a coarse surface because large grains are
formed which do not touch, and the thickness must be
increased before conductivity is observed. Such films tend
to be unstable because their conductivity depends upon
contacts between large grains.
The most suitable substrate materials are found amongst
glasses and ceramics. Good results have already been
obtained using glasses of high silica content such as Pyrex
or Vycor. These are two of the few glasses unaffected by
water vapour. Many other glasses, including some boro-
silicates, devitrify in contact with water, and their surfaces
become powdery because small crystals of metal silicates are
formed. Soda-lime glasses are not used because their sur-
faces are also chemically unstable. During flame polishing,
when the glass is bombarded by thermally produced gas ions
or ionic bombardment in a glow discharge, free sodium ions
are active at the surface of the glass. They combine with
Water vapour from the gas atmosphere to form sodium
hydroxide and deliquescent sodium silicate by reaction with
VOL. 12, DECEMBER 1961 669
silica in the glass. Some further reaction with the deposited
film must be expected.
Ceramics possessing good chemical and mechanical
properties are available; however, their surface smoothness
is often variable because of the sintering process used in
their manufacture. Glazing is not always a satisfactory
solution to this problem because standard glazes are often
based upon some of the unsuitable glasses already described.
Very careful examination of surface smoothness is needed
when choosing a ceramic material for the support of vacuum-
deposited films.
The temperature coefficient of linear expansion of metals
is usually an order higher than that of glass or ceramics, and
this factor partly contributes t o the high intemal stresses
which have been observed in thin films. However, once the
films have been annealed, the effect of the expansion of the
base on the resistance of the film is very small, compared with
the average temperature coefficient of resistance of films, and
is insignificant when compared with that of bulk metals.
The resistance alloy
In the early stages of deposition of a metal film, aggregates
of metal atoms (nuclei) are formed on the substrate. The
number of nuclei is dependent upon the physical and chemical
properties of the metal and substrate, and upon the rate of
deposition. As the nuclei increase in size they grow together,
eventually to form a continuous film. The second stage of
growth is marked by the onset of electrical conductivity, and
the rate of change of resistance with film thickness is very
high. Unfortunately, the most useful resistance values
coincide with this unstable region of thickness for many
metallic conductors.
The most successful high-resistance films have been made
by depositing chromium and alloys of chromium with nickel,
silicon, titanium, etc. For example, nickel-chromium alloys
have a high bulk resistivity (80-130 pQ cm) and therefore
films of this alloy are much thicker than films of the pure
metals for the same resistance value. Films of resistance
400 Q/square are at least 80 A thick, more or less continuous
and are outside the very unstable region of thickness.
Nickel-chromium alloys also have a low temperature coeffi-
cient of resistance in bulk, and are very resistant to chemical
attack because of the compact protective oxide layer which
forms in contact with an oxidizing atmosphere. The forma-
tion of double oxides having a spinel structure has been
shown on nickel-chromium alloys under examination by
electron diffraction, and this reason has been given to account
for their high chemical stability.
Evaporation conditions
The lowest values of temperature coefficient of resistance
and the best stability are achieved in films deposited under
conditions favouring oxidation. During deposition the
substrate is heated to relieve the iniemal stress in the film,
but this treatment can also increase the rate of oxidation.
The residual gas atmosphere in the chamber of a kinetic
vacuum system is highly oxidizing due t o the high proportion
of water vapour at the normal working pressure ( 10e4mmHg).
Assuming that the partial pressure. of water vapour is only
mm Hg, then it is calculated approximately that
5 x 10" molecuies cm-2s-1 strike the substrate surface.
If the rate of deposition of chromium metal is about 3 A s-',
then ten water vapour molecules strike the surface for every
BRITISH JCURNAL OF APTLIED PHYSiCS