The range of infrared wavelengths is from 0.76 um to 1000 um. It can be divided into four groups which are near infrared (0.76 um~1.5 um), middle infrared (1.5 um~5.6 um), far infrared (5.6 um~25 um), and extra far infrared (25 um~1000 um). Crawford F.J.
[1] presented the spectral dependence of radiation on temperature shown as Figure 2-1-1 Every object that is not at absolute zero emits and reflects electromagnetic radiation. The wavelength of the radiation is a function of the temperature of the object. Shorts wavelengths are emitted by higher temperatures and longer wavelengths are emitted by cooler objects. The short wavelengths between 0.45 um to 1um used by visible cameras are emitted by very hot sources, such as the sun or incandescent light bulbs. However, the longer infrared wavelengths between 3 and 14 um are emitted by objects at temperatures around 300 K or 25 oC.
Figure 2-1-1 Spectral dependence of radiation on temperature
Radiation is either scattered or absorbed by gas molecules, rain, and snow as it propagates through the atmosphere. Figure 2-1-2 shows the atmospheric transmission characteristics from visible to 14 um. It can be noted that the two infrared transmission windows at 3 to 5 um and 8 to 12 um are the wavelengths encompassed by terrestrial temperatures. These transmission windows dictate the choice of wavelengths used in the infrared sensor design.
Figure 2-1-2 Atmospheric transmission characteristics 1.2 Infrared (IR) detector
Over the past several years, uncooled infrared (IR) radiation detectors have been rapidly developed into a large size of focal plane arrays (FPA). IR detectors have been an important technology for both military and civilian application, such as night vision, surveillance, detection of gas leakage, fire rescue operation, manufacturing quality control, early fire detection, and missile tracking, guidance, discrimination, and interception [2-5]. There are basically two types of IR detectors: photon and thermal detectors [6]. Traditionally, photon detectors are preferred primarily due to their superior
sensitivity. However, the photon detectors must be cooled to approximately the temperature of liquid nitrogen, with the cooling apparatus typically which is the most expensive component in a photon detector’s IR camera. Most thermal detectors, on the other hand, don’t need such an apparatus, and they are uncooled and comparatively inexpensive. In addition, the recent advances in micromachining technology have made it possible to fabricate highly sensitive thermal IR detectors. Hence it can decrease the cost of the system, offer improved reliability and mean-time-before failure (MTBF), instant operation, sensitivity in the spectral region of 8 um to 12 um. Conventional uncooled IR detectors, which have long been used for human image detection and temperature measurements, have a very small number of elements, sometimes only one. This is insufficient for tracking moving IR images; here, it is necessary to increase the number of elements and to integrate them into a two dimensional (2 D) array. Such an array sensor is called a focal plane array (FPA), or an array image sensor. Three different types of uncooled FPA have been developed. They are bolometer detectors [3, 7], thermopile detectors [8], and pyroelectric detectors [9-10]. Bolometers are generally easier to fabricate than pyroelectric detectors and have better responsibility than thermopiles [6].
The bolometer detectors are adopted in this dissertation.
1.3 Micromachining technology with integrated circuits for micro-sensor applications
In recent years, microelectromechanical systems (MEMS) have emerged as a very promising field of researches and applications. There is a tendency towards developing a smart sensor which is a combination of sensors and integrated circuits (ICs) in the same chip. Practical implementation of smart sensors has still a lot of problems to be solved such as the difficulty in integrating different fabrication technologies.
From the very beginning, integration of microelectromechanical systems (MEMS) with integrated circuits (ICs) was a major attraction of silicon micromachining
technology. Practical implementation has not been easy. While the first pressure sensors reached the market in the 1960s, many first integration efforts failed, yielding the first practical, integrated pressure sensors in the 1980s [11]. The 1990s brought multiple product launches of integrated MEMS-IC devices, such as acceleration sensors, ink jet print heads, and display chips. In the current decade, integrated gyro sensors entered the market. All these efforts focused on integrating MEMS processes and IC processes on the same wafer. Incompatibilities forced long and expensive development cycles. Several approaches of process integration between IC and MEMS were developed over the past two decades. Generically, they classify into three categories as the following:
(a) Integration of MEMS on Top of IC
(b) Lateral (side-by-side) MEMS and IC integration (c) Vertical wafer-level MEMS-IC integration.
¾ Integration of MEMS on Top of IC
The most straightforward method of integration is to build the MEMS device directly on top of the CMOS wafers. This has the severe disadvantage of requiring strict process compatibility. MEMS structures are limited to those that can be surface micromachined within the CMOS thermal budget (<400 ◦C). It rules out LPCVD polysilicon and silicon fusion bonding, a staple for many MEMS devices. Another restriction is that the exposed materials, a low-temperature oxide and aluminum, limit the chemical means available for processing.
¾ Lateral (Side-by-Side) MEMS and IC Integration
This approach overcomes some process incompatibilities between MEMS and CMOS. This method fabricates any CMOS incompatible processes first and then can do both bulk and surface micromachining.
¾ Vertical Wafer-Level MEMS-IC Integration
This approach bonds together two or more wafers. At least one wafer is MEMS and at least one other is CMOS. Each is fabricated in a dedicated foundry. Vertical wafer level integration of circuits and MEMS has several advantages over integrating MEMS
and active circuitry at the process level:
(1) It lacks the restriction of process compatibility.
(2) It does not suffer the real estate penalty and inefficiency of lateral integration techniques.
(3) Because the MEMS and CMOS are fabricated separately prior to integration, it affords the designer absolute flexibility in the choice of active circuit process options. High-density digital, mixed signal, high voltage (HV), BiCMOS, and RF can all be integrated using the same process steps.
The example of the wafer-level MEMS and IC integration is shown in Figure 2-1-3 [11]. Two wafers are preprocessed to fabricate the bond metallization. The bonding recipe requires a precise bonding pressure and temperature profile. It is possible to create bonds which are purely structural, or which can also conduct signals between the MEMS and the CMOS, thus functioning as a part of the electrical circuit.
Figure 2-1-3 Vertical wafer-level MEMS-IC integration
1.4 Surface micromachining on top of IC for bolometer detector applications
Although the technology of vertical wafer-level MEMS-IC integration provides many advantages over MEMS on top of IC and lateral MEMS-IC integration, it may not be the best choice for the IRFPA applications, which require a large number of elements and are very susceptible to environmental thermal noise and electrical noise. In order to obtain the best device performance of IR detectors, the surface micromachining on top of IC is the promising candidate for infrared detector applications.
Hence, a low-thermal budget (<400℃), CMOS-process compatible, surface micromachining process becomes more and more attractive, and it is widely applied to commercial products and consumer goods due to its benefit of easy integration with ICs.
In addition, a low temperature (<400℃) process makes MEMS sensors be able to fabricate directly on ICs as a post-process. This is a very convenient approach to integrate MEMS devices with driving and sensing circuits into a chip, and this integration can improve immunity against noise and increase sensing sensitivity. The primary structures of the surface micromachining are suspension structures such as cantilever beam, bridge, diaphragm, and membrane [12]-[15]. In these structures, flatness of a freestanding diaphragm or a cantilever beam is an important issue for optical applications which are like optical switches and micro-mirrors [15]. However, the flatness and long-legs of membranes are even more a challenge than other devices to design and fabricate for thermally isolated applications, especially as low cost uncooled IR (infrared) microbolometers [16], [17].
The IR microbolometers are radiation sensors with an infrared absorber. Under IR illumination, the temperature of the absorbing layer increases and then the resistivity of the sensitive material changes. In order to acquire a microbolometer with high responsivity, the thermally isolated structure of the device is vital to ensure a maximum increase of temperature due to the absorption of IR radiation [16]. The most important part of the whole microbolometer is the suspension membrane serving as a structural layer on which the sensing material and the signal metal lines are located. Besides, regarding thermal isolation of the microbolometer, many microfabrication approaches have been presented to reduce the thermal losses and thermal mass of the membranes of the microbolometer structures. These approaches contain the use of low thermal conductivity materials and thermally isolated structures obtained by surface or bulk micromachining techniques [17]. In order to integrate easily with CMOS sensing circuits, the surface micromachining technique is preferable due to its small feature size and CMOS compatible processes. However, in the surface micromachining technique, the sticking effect and residual stress play an important role in determining whether the
microstructures are suspended or collapse during the release process. In general, the released microstructures are apt to be attached to the underlying layers. There have been many studies about sticking effect and residual stress with regard to the cantilevered beam structures by developing several different models [18]-[19]. However, no research has been done on the influence of the anchor profile of the membrane microstructure with residual stress. In addition, during the release process, drying of the delicate microstructures is another important issue for whether the device is successful or failure.
Many release methods have been presented and compared [20]. To date, most drying methods of release apply CO2 supercritical point to overcome the sticking effect, but this method leads to low throughput, high cost and expensive apparatus. In this dissertation, we used hot plate to dry the release-etch microstructure to substitute for CO2 supercritical point method. Although Takeshi [21] described the effects of elevated temperature treatments in microstructure release procedures, these effects were only been investigated for cantilever.
1.5 In our research
We first develop the test microbolometer structure from two aspects including structural design and material selection to obtain high thermal isolation structures. Then, the structure fabrication has been developed to reach the aims of lower processing cost, higher fabricating reliability, and compatible with CMOS processes. The design of the test microstructure by controlling the anchor profile containing sidewall conformal factor (SCF) and sidewall angle is investigated. In addition, the temperature effect during release procedures for single membrane and membrane array is also discussed.
Chapter 2 Bolometer Operation and Theory
2.1 Bolometer operation
In general, the IR microbolometers are radiation sensors with an infrared absorber.
Under IR illumination, the temperature of the absorbing layer increases and then the resistivity of the sensitive material changes. In this research, the proposed pixel structure of a microbolometer is presented in Figure 2-2-1. The operation of the device is based on three steps as the following:
(1) Distant objects that are not at absolute zero temperature emit thermal radiation and impact on the thermal absorbing layer of a microbolometer device. This is the first step that optical signal is converted into heat flux.
(2) The absorbed thermal energy increases the temperature of the sensing layer. This is the second step that heat flux is converted into temperature difference.
(3) The increasing temperature results in the resistance change of the sensing layer.
This is the third step that temperature difference is converted into electrical signals.
In order to obtain the largest temperature difference on the basis of these three steps, the absorbed efficiency and the thermal isolation of the microbolometer device must keep as good as possible.
Figure 2-2-1 Pixel structure of a microbolometer
For the sake of making bolometer operation more clear, we further consider the input bias and heat-sink phenomenon [22]. A bolometer measures changes in the heat input from the surroundings and converts this into a measurable quantity such as a voltage or current. A bolometer therefore typically consists of an absorber and a thermometer of heat capacity C, connected by a small thermal conductance G, to a heat sink held at a fixed temperature T0 (it is usually 300 K)as demonstrated in Figure 2-2-2(a).
The energy P of the incident radiation is converted into heat in the absorber, leading to a temperature rise∆T =T−T0 =P/C, until the radiation power flowing into the absorber is equal to the power flowing into the heat sink through the weak thermal link. The temperature rise is subsequently measured and is directly proportional to the deposited energy.
Figure 2-2-2 (a) Schematic illustration of bolometer operation (b) Bolometer bias circuit
The bolometer is in general placed in a bias circuit with a voltage source and load resistor as shown in Figure 2-2-2(b). A constant current I, generated from the load resistor
RL and bias voltage Vbias, flows through the bolometer. Provided this bias power ( Pbias = VbiasI) remains constant, the incoming signal power ( Psignal), collected by the detector, will cause the bolometer temperature T to increase according to:
G P P
T
T = 0 +( signal+ bias)/ (2-1)
The temperature rise causes a change in the resistance of the bolometer and consequently in the voltage across it. This change in voltage is amplified and measured.
The thermometer is therefore made of a material that ideally has a large change in resistivity for a small change in temperature.
2.2 General performance concepts [23]
The important figure of merit for a bolometer is the sensitivity (S), detectivity (D), noise equivalent power (NEP) and time constant (τ ). The NEP is a measure of the sensitivity of a bolometer, and is defined as the power absorbed that produces a signal to noise of unity at the output. NEP can also be defined as the minimum incident radiation power necessary to obtain a signal to noise of unity.
Radiation incident on a solid state device will cause the absorption of photons and the creation of lattice vibrations or phonons. This will act to increase the resistivity of a solid state device. Two parameters are convenient to define before stating an expression for a bolometer’s votage signal. The first is the heat conductance of G defined as dP/dT where P is the radiative power lost or absorbed by the device and T is temperature. The second parameter is the temperature coefficient of resistance, α, where α = (1/R)dR/dT is the fractional change in resistance per degree Kelvin. We now assume the device is driven by a constant current I, and is enclosed in a chamber at temperature T0. Since the current through the resistor heats it slightly, we define the temperature of the detector element as
T and assume (T-T0) << T0. The signal voltage becomes It follows that bolometer sensitivity, S=VS/Ps, is given by
G R S = Iα
(2-3)
For high sensitivity, one requires a large electrical resistance, a large current, a large temperature coefficient of resistance and a small heat conductance. In order to obtain a large temperature coefficient of resistance is accomplished by choosing the proper material. The bolometer’s resistance is controlled by maintaining a small film thickness and a large length to width ratio. The latter requirement, however, conflicts the need for resistance uniformity across an array of bolometers for infrared video imaging.
In such cases, it may be necessary to sacrifice the benefits of a large resistance.
Minimizing the heat conduction is probably the most interesting aspect of bolometer design. The large thermal conductance provided by a bulk substrate will draw heat from the temperature sensing element too fast and lower the sensitivity. A large thermal conductance, however, may be desirable in the construction of very high speed bolometer if a reduction in sensitivity is tolerable. By quickly drawing the excess heat from the bolometer resistance element, fast changes in temperature or infrared images can be detected.
To quantify the speed of a bolometer, consider a detector element with heat capacity C defined as C = dE/dT where E is the total internal energy of the element. The bolometer time constant is then τ = C/G in analogy to a single time constant circuit. It should be noted that optimizing the sensitivity by lowering G and optimizing the speed by increasing G cannot be done without simultaneously lowering the heat capacity.
Intuitively, the need for a small C by considering that a smaller change in the bolometer internal energy is needed to create an appreciable change in temperature. Therefore a
small incident energy can be used to cause a change in the voltage signal.
Equation (2-3) can be further modified to include the effect of illumination chopping frequency, ω, as the following:
A factor, η, has also added to account for incomplete optical absorption by the absorbing layer. The factor is called the coefficient of absorptance.
In terms of noise figure for a bolometer, the three dominant noise factors must be considered. All solids at a finite temperature suffer from temperature fluctuations and Johnson noise. The third noise source arises from the electrical noise in the amplifier electronics. The mean-square fluctuation in the voltage signal per unit of bandwidth is obtained by summing the square of the two independent factors. The first factor is the amplifier noise and the second factor arises from the temperature fluctuations in the bolometer resistive element.
2
Where k is Boltzmann’s constant and TN is the effective input noise temperature of the amplifier. Because the body has a conductance G and capacity C with time constant τ, we know from basic network theory that the power spectrum of such a body is of the form 2 2 2
where K is a constant of proportionality found by considering the entire temperature fluctuation over all frequencies. Besides, from Boltzmann’s relationship, the probability of the resistive element has energy Ei and can be expressed
p(Ei)=Ae−Ei/kT with
∑
− /= 1
From the definition of heat capacity,
2 2 2 12 2 Combining the above with equation (2-6), we have,
∞
∫
+ = Solving for K yields the noise spectral density,) For frequencies much less than 1/τ, the final temperature fluctuation of the sensor, per unit of bandwidth, is For a body in thermal equilibrium, one-half of the fluctuation is due to emission of radiation while the other half is due to absorption. It follows that the mean-square temperature fluctuation per unit of bandwidth is
4 (1 ) The resultant voltage fluctuation is
4 (1 )
The NEP is defined as the minimum incident radiation power, PS, necessary to produce S/N = 1. From the latter equation we therefore have the noise equivalent power in units of watts per Hz1/2,
The second term indicated an increasing NEP with bias current. The third term is the Johnson noise and indicates a decreasing NEP with bias current, associated with the increased signal response. It can be written in terms of the sensitivity, using equation (2-2),
The specific detectivity, D*, is defined as
NEP BW
D A( )
*= (2-20)
Where A is the device area and BW is the bandwidth of the noise power. The detectivity D* is less sensitive to the detector area A.
2.3 Analysis of heat transfer mechanisms for bolometer structures
The main structure of the microbolometer consists of two parts; one is the thin film resistor, another is the membrane structure. The membrane structure offers a excellent thermal isolation and supports the whole device with two narrow legs. The basic principle of microbolometer is that raising material temperature by absorbing IR radiant energy, and when temperature changes, the electrical resistance will be change. The heat balance
equation is expressed by mathematically as follows:
equation is expressed by mathematically as follows: