Microhotplate

During the last years, so-called microhotplates (pHP) have been developed in order to shrink the overall dimensions and to reduce the thermal mass of metal-oxide gas sensors [7,9,15]. Microhotplates consist of a thermally isolated stage with a heater structure, a temperature sensor and a set of contact electrodes for the sensitive layer. By using such microstructures, high operation temperatures can be reached at comparably low power consumption (< 100 mW). Moreover, small time constants on the order of 10 ms enable applying temperature modulation techniques with the aim to improve sensor selectivity and sensitivity. 

The development of microhotplates is strongly coupled to novel nanotechnological fabrication strategies as well as microdeposition and microstructuring techniques for the respective metal oxides. 

To date, most microhotplate-based chemical sensors have been reahzed as multichip solutions with separate transducer and electronics chips [16-19]. The co-inte-... 

Microhotplates, however, are not only used for metal-oxide-based gas sensor applications. In all cases, in which elevated temperatures are required, or thermal decoupling from the bulk substrate is necessary, microhotplate-like structures can be used with various materials and detector configurations [25]. Examples include polymer-based capacitive sensors [26], pellistors [27-29], GasFETs [30,31], sensors based on changes in thermal conductivity [32], or devices that rely on metal films [33,34]. Only microhotplates for chemoresistive metal-oxide materials will be further detailed here. The relevant design considerations will be addressed. 

The basic components of a microhotplate-based sensor system are shown in the lower part of Fig. 2.1. They include the microhotplate and dedicated sensor electronics. As shown in Fig. 2.1 the electronics part may feature a temperature controller, read-out and measurement circuitry for the different sensor elements, or a first data processing stage. Another interesting feature is an embedded sensor interface. 

In the upper part of Fig. 2.1, three main blocks are shown that represent all the issues related to the design and layout of microhotplate-based sensors and electronics. They include ... 

The third block in Fig. 2.1 shows the various possible sensing modes. The basic operation mode of a micromachined metal-oxide sensor is the measurement of the resistance or impedance [69] of the sensitive layer at constant temperature. A well-known problem of metal-oxide-based sensors is their lack of selectivity. Additional information on the interaction of analyte and sensitive layer may lead to better gas discrimination. Micromachined sensors exhibit a low thermal time constant, which can be used to advantage by applying temperature-modulation techniques. The gas/oxide interaction characteristics and dynamics are observable in the measured sensor resistance. Various temperature modulation methods have been explored. The first method relies on a train of rectangular temperature pulses at variable temperature step heights [70-72]. This method was further developed to find optimized modulation curves [73]. Sinusoidal temperature modulation also has been applied, and the data were evaluated by Fourier transformation [75]. Another idea included the simultaneous measurement of the resistive and calorimetric microhotplate response by additionally monitoring the change in the heater resistance upon gas exposure [74-76]. 

Most microhotplate-based chemical sensors have been realized as multi-chip solutions with separate transducer and electronics chips. One example includes a gas sensor based on a thin metal film [16]. Another example is a hybrid sensor system comprising a tin-oxide-coated microhotplate, an alcohol sensor, a humidity sensor and a corresponding ASIC chip (Application Specific Integrated Circuit) [17]. More recent developments include an interface-circuit chip for metal oxide gas sensors and the conccept for an on-chip driving circuitry architecture of a gas sensor array [18,19]. 

The first monolithic devices have been presented at the same time by a group at NIST and a group at the Physical Electronics Laboratory (PEL) of ETH Zurich [77-81]. The NIST chip hosts an array of microhotplates integrated with transistor switches and a readout amplifier for the sensitive layer. The device presented by PEL includes an analog temperature controller and a logarithmic converter for reading out the sensor values. This was the first monolithic realization of an embedded system architecture with integrated microhotplate. 

A cross-sectional schematic of a monolithic gas sensor system featuring a microhotplate is shown in Fig. 2.2. Its fabrication relies on an industrial CMOS-process with subsequent micromachining steps. Diverse thin-film layers, which can be used for electrical insulation and passivation, are available in the CMOS-process. They are denoted dielectric layers and include several silicon-oxide layers such as the thermal field oxide, the contact oxide and the intermetal oxide as well as a silicon-nitride layer that serves as passivation. All these materials exhibit a characteristically low thermal conductivity, so that a membrane, which consists of only the dielectric layers, provides excellent thermal insulation between the bulk-silicon chip and a heated area. The heated area features a resistive heater, a temperature sensor, and the electrodes that contact the deposited sensitive metal oxide. An additional temperature sensor is integrated close to the circuitry on the bulk chip to monitor the overall chip temperature. The membrane is released by etching away the silicon underneath the dielectric layers. Depending on the micromachining procedure, it is possible to leave a silicon island underneath the heated area. Such an island can serve as a heat spreader and also mechanically stabihzes the membrane. The fabrication process will be explained in more detail in Chap 4. 

The sensitive layers that have been used throughout this book were produced and deposited by AppliedSensor (AS, Reutlingen, Germany). Although metal-oxide-based microhotplate-sensors are already commercially available, a brief description of the paste production is given for the sake of completeness. The process is detailed in [82]. 

Another issue is how to include heat conduction and dissipation through ambient air. A heat transfer coefficient, h, is commonly used in 2-dimensional simulations, which is difficult to determine, since it is strongly depending on the package of the microhotplate sensor. It is, therefore, in most cases introduced as a parameter that is varied to fit the experimental data [45,94]. 

A key issue in monolithic-system design is the simulation of the microhotplate coupled to the circuitry to ensure full chip functionahty. This requires an adequate description of the microhotplate and an implementation in a language that is applicable to circuitry simulations. 

Fig. 3.1. Schematic of the modeUing steps needed to come from a microhotplate layout to a model description for system-level simulations...

The suggested procedure to arrive at this goal is presented in Fig. 3.1. It starts with the transfer of a certain microhotplate layout into a geometry model for a complex FEM simulation. This step is shown in Fig. 3.2 and will be explained in more detail in one of the next sections. A complex 3-d FEM simulation is then performed. The results of this simulation are used to produce a lumped-element model. This model is translated into a hardware description language (HDL). Using the resistances of the device elements such as the heater resistance, Rheat> and the resistance of the temperature sensor, Rx. co-simulations with the circuitry can be performed. 

The aim in converting the microhotplate into a geometry model for the FEM simulation was to find a model that is as simple as possible but includes all relevant processes. The model assumptions to be explained in detail in the following section and the steps to arrive at the model are represented in fig. 3.2. The feature on the bottom left-hand side represents the microhotplate schematic. The microhotplate exhibits a symmetric design so that a simulation of one quarter is adequate. Geomet-... 

Fig. 3.2. Schematic showing the translation of a microhotplate layout into a geometry model description for FEM simulations...

A possible heat-loss mechanism includes thermal radiation, pjad- The hotplate operating temperature range is up to 350 °C, for which radiation losses are considered to be negligible [94,96]. In case of higher temperatures, radiation losses would have to be included [97,98]. The overall loss owing to radiation scales with the total heated area. A rough estimate for radiation losses of the presented microhotplate at 300 °C is 2% of the overall hotplate power consumption. 

A characteristic measurement includes the determination of the microhotplate temperature as a function of its power consumption. The curves can be fitted by a second-order polynomial using the coefficients tjo and t]i. 

The thermal resistance will be temperature-dependent as canbe seen in Eq. (3.24), which is not only a consequence of the temperature dependence of the thermal heat conduction coefficients. The measured membrane temperature, Tm, is related to the location of the temperature sensor, so that the temperature distribution across the heated area will also influence the thermal resistance value. The nonlinearity in Eq. (3.24) is, nevertheless, small. The expression thermal resistance consequently often refers to the coefficient t]o only, which is used as a figure of merit and corresponds, according to Eqs. (3.24) and (3.25), to the thermal resistance or thermal efficiency of the microhotplate at ambient temperature, Tq. The temperature Tm can be determined from simulations with distinct heating powers. The thermal resistance then can be extracted from these data. 

---