Advancements in
infra-red array
detectors
Gareth J. Monkman
Infra-red detectors can be either quantum or thermal devices. The former are usually much more sensitive but only to a limited wave-length. The latter lack sensitivity but exhibit very broad bandwidths extending from the ultraviolet far into the infra-red (Monkman, 1996). Thus, when measurement and analysis of a single radiation source are all that is required, the necessary spectral width may be covered by employing a range of discrete sensor elements. Unfortunately this approach is unsatisfactory where the acquisition of complete infra-red images is required. Recent advances in micromachine technology have resulted in sensitivity improvements for far infra-red thermal detectors to a degree where they may now be applicable to single and two-dimensional array sensors suitable for thermal imaging applications.
2. Quantum detectors
When designed as quantum detectors, the band gap energy of most semiconductor materials determines the maximum usable wavelength at which optical sensors are capable of operating. Formula (1) describes the relationship between the band gap energy Egin electron volts and maximum usable
wavelength in micrometreslmax.
max 1:24
Eg
m 1
The band gap energies for a selection of semiconductor materials commonly used for optical detectors are shown in Table I (Bar-Lev, 1979).
A quick calculation using equation (1) in conjunction with Table I will reveal, even for the best material listed, a maximum measur-able wavelength of about 3.4m. Though adequate for simple non-contact temperature measurement, high accuracy-systems and most infra-red imaging devices require good sensitivity up to a range of at least 8m and when used in spectrometers must be capable of covering an even wider spectrum, prefer-ably as far as 24m.
Although the manufacture of visible and near infra-red wavelength semiconductor detector arrays is a well established art, no array devices of this nature are capable of
The author
Gareth J. Monkmanis a Professor at Fachhochschule Regensburg, Regensburg, Germany.
Keywords
Infra-red, Sensors, Thermal imaging, Thermocouples
Abstract
Despite many advances during the last decade in both infra-red sensor and solid state camera technology, until now little headway has been made in the production of cost-effective semiconductor sensor arrays capable of operating far into the infra-red. Old ideas, renewed by the capabilities offered by the latest micromachine technol-ogy, may change all this. Reviews the problems associated with building such sensor arrays before introducing some interesting new research results.
Electronic access
The research register for this journal is available at
http://www2.mcb.co.uk/mcbrr/sr.asp
The current issue and full text archive of this journal is available athttp://www.emerald-library.com
Sensor Review
Volume 19 . Number 4 . 1999 . pp. 273±277
operating into the far infra-red. Most high quality thermal imagers rely on thermionic vidicon type tubes or employ some form of mechanical scanner to project the radiation directly onto a single thermal sensor element. Both solutions are relatively expensive and in many cases the detector element must be cooled with liquid nitrogen, or at least be maintained at a stable and lower than ambient temperature.
3. Thermal detectors
The alternatives to quantum detectors is of course thermal sensors such as bolometers, thermocouples and ferroelectric pyrometers. These all enjoy a wavelength-independent response but the sensitivity is extremely poor in comparison with similar sized semi-conductor elements. Both the sensitivity and the time constant of these detectors are proportional to element size making trade-offs between sensitivity and speed inevitable. Additional problems arise when a multiplicity of such detectors are to be linked into 1D or 2D arrays. In this case the surface area of individual elements and their distance apart must be minimised. This limits the surface area proportional sensitivity and because of the thermal conductivity of the detector material and the substrate upon which they are mounted, cross-talk levels generated in similar configurations of semiconductor quantum detectors are a problem.
The simplest form of sensitivity measure-ment is theresponsivity. This is defined as the amount of electrical output from the detector per unit of illumination power. It may be defined as current responsivity RI[A/W] or
voltage responsivity Rv[V/W]. One additional
problem which plagues all infra-red sensors is ``noise''. Any detectable signal must be at least as high as the noise voltage Vn(or noise
current In). The noise equivalent power
(NEP) is the smallest detectable change in radiation capable of generating this voltage change in the detector output and the detectivityD is the reciprocal of the NEP. This is often expressed as the specific detectivity D* and is related to D by the sensor surface
area A, and the bandwidth Df, equation (2):
D DpAf mpHzWÿ1 2
or expressed in terms of the responsivity:
D
Clearly, the physical dimensions of the sensing element, including the surface area play an important role.
3.1 Thermopiles
Thermocouples are junctions formed between two dissimilar materials, usually metal alloys or semiconductors, across which a tempera-ture proportional voltage is produced ± Seebeck effect. The usual manner of increas-ing sensitivity is to cluster a large number of elements closely together ± known as thermopiles. The combined output voltage V is given by equation (4).
VN 1ÿ2 T 4
Here N is the number of thermocouple junctions andDT is the temperature differ-ence between a referdiffer-ence junction and the measurement junction(s). Certain semi-conductor materials have very high Seebeck voltages, a selection of examples of which is given in Table II. These may be compared with the ideal case of single crystal silicon with a carrier concentration of around 1019cm±3 (Van Herwaarden and Sarro, 1986).
For example, a thermopile comprising ten Bi0.87Sb0.13/Bi0.5Sb1.5Te3 thermocouples
would produce a theoretical output voltage of 3.3mV/K ± an easily measurable quantity. On this basis, a team at the Institute for High Technology Physics, in Jena, have developed 256 pixel arrays suitable for direct integration with on-chip signal processing. Each pixel consists of ten tiny thermocouples, each micromachined to a thickness of 0.4mm, serially connected to form a thermopile (GuÈttichet al., 1999).
The results shown in Table III clearly demonstrate the problems associated with thermal conductivity of the surrounding materials. With the excellent thermal isolation provided under vacuum conditions the detectivity is very high but the response time
Table IWavelength dependence on band-gap energy for semiconductors
Material Si Ge GaP GaAs GaSb InAs SiO2
long. The thermal time constant is signifi-cantly reduced in the nitrogen (at normal pressure) atmosphere but at the cost of sensitivity.
3.2 Ferroelectric pyrometers
For ferroelectric pyrometers, the general voltage responsivity is given by equation (5):
Rv
e is the proportion of crystal structure which is thermalised (%);
ois the angular frequency of modulation [S±1];
A is the exposed crystal surface area [m2]; p is the pyroelectric factor
[C m±2K±1];
R is the internal resistance (including that of the measurement device)
[V A±1];
G is the thermal conductivity of the crystal [J s±1K±1kg±1];
tTis the thermal time constant [s];
tEis the electrical time constant [s].
The parameters G and p are usually material dependent constants over which we have little control. Increasing the surface area, A gives an improvement in sensitivity provided the whole surface is illuminated; otherwise e is reduced. However, as already explained, for array detectors, A must be physically as small as possible.
As capacitive elements, ferroelectric crystals react only to changes in radiation intensity. Consequently ocannot be zero and so the radiation falling on them must be modulated. This is normally achieved by means of a
mechanical chopper. The thermal and elec-trical time constants are likely to be in the millisecond range. So long as the modulation provided by mechanical means is only a few Hz then the modulation period is not likely to approach that of the time constants. Given such a relationship between modulation frequency and time constants it can be seen from equation (5) that quartic products in the denominator can be neglected. This is the normal situation for pyrometric sensors allowing the expression for responsivity to be reduced to equation (6):
Rv
e A p R
c V=W 6
where c is the heat capacity [ J kg±1K±1] of the crystal material.
With modern CMOS technology, the input resistance to any measuring device is extremely high. If R in expression (6) represents only the internal resistance of the crystal then R can be replaced by the resistivityr[Om] of the material (a constant) divided by the thickness of the crystal element d to yield equation (7):
Rv
e A p
c d V=W 7
Now clearly, if the thickness of the crystal layer d can be minimised then there are substantial gains to be made.
Collaboration between the TU Dresden and local firm DIAS GmbH has, through a series of mechanical and chemical polishing processes, reduced the thickness of lithium-tantalate (LiTaO3) to around 20mm.
Additional ion-beam etching has been used to reduce this dimension further to a final thickness of 2mm. Figure 1 shows the effects
Table IISemiconductor thermocouple materials
Material Bi (metal) Sb (metal) Bi0.5Sb1.5Te3 Bi0.87Sb0.13 Si (ideal)
ai(mV/K) ±68 51 230 ±100 ±450
Table IIITwo designs of micromachined thermopile
on responsivity from reduced thickness. Linear arrays of up to 256 elements have been produced for which very high element responsivities of 500,000 V/W are claimed (Norkuset al., 1999). Figure 2 shows for this same device an electron microscope picture of the electrical bonding between the pyrometer elements on the left and the signal processing on the right. Figure 3 is an overall view of the 128-element array.
Another research group, at the Ecole Polytechnique FeÂdeÂrale Lausanne have also developed ceramic thin film detectors, intended for infra-red gas spectrometry. Using sol-gel techniques a lower thickness of 1mm using Pb(Zr, Zi)O3 has been attained.
The more usual responsivity of 500V/W is claimed but the low noise level gives a detectivity of 160cm6106cm (Hz/W which is comparable with the thermoelements pre-viously described). Again, thermal cross-talk is a problem with figures typically around 20 per cent between elements (Willing, 1999). Such applications require only mirror focus-ing techniques, thus avoidfocus-ing the need for expensive lenses.
Owing to the very small current flow from ferroelectric elements ± typically in the pico-ampere range, either closely integrated or on-chip signal processing is essential. In most cases digitisation immediately follows making computer integration via parallel or serial interfaces relatively simple.
4. Conclusions
Micromachine technology clearly makes the production of infra-red sensor elements so small and thin that the overall sensitivity is not necessarily compromised by the improvements in response time. Further-more, the solid state approach does not only raise hopes of finally replacing conventional tubes or expensive mechanisms but also allows for integration with on-chip signal processing ± a significant factor in overall system cost reduction. Unfortunately, one major problem remains for all wide band infra-red camera systems used for thermal
Figure 1Dramatic increase in responsivity for reduced element thickness
imaging in that the optics can represent up to 40 per cent of the total cost. Focusing radiation at the far infra-red end of the spectrum cannot be achieved with the same types of glass used at visible wavelengths. Pure germanium is the standard material and unless a cheaper substitute can be found, thermal imagers will remain expensive pieces of equipment for some time to come.
References
Bar-Lev. A. (1979),Semiconductors and Electronic Devices, Prentice-Hall, London.
GuÈttich. R., Baier, V., Dillner, U., Kessler, E., MuÈller, J., Berger, A., Behrendt, D. and Suphan, K-H. (1999),
``Thermoelectric IR radiation linear arrays with hybrid-integrated signal processing'',Proceedings 9th International Conference, Sensors, Transducers & Systems, NuÈrnberg, pp. 29-34.
Monkman, G.J. (1996), ``Industrial infra-red sensors'', Sensor Review, Vol. 16 No. 3, pp. 26-9. Norkus, V., Gerlach, G. and Hofmann, G. (1999),
``Uncooled linear arrays based on LiTa03'', Pro-ceedings 9th International Conference, Sensors, Transducers and Systems, NuÈrnberg, May, pp. 23-8. Van Herwaarden, A.W. and Sarro, P.M. (1986), ``Thermal sensors based on the Seebeck effect'',Sensors and Actuators, No. 10, pp. 321-46.
Willing, B. (1999), ``Infra-red gas spectrometry based on a pyroelectric thin film array detector'',Proceedings 9th International Conference Sensors, Transducers and Systems, NuÈrnberg, May, pp. 149-54.
