Abstract
When an ionizing particle passes through a semiconductor device, it transfers energy and generates electron-hole pairs along its path. The excess carriers are subsequently transported throughout the semiconductor’s volume via ambipolar diffusion until they either recombine or are collected and extracted typically by a biased contact or a reverse-biased p-n junction. To predict the transient electrical behavior of complementary metal-oxide semiconductor (CMOS) devices and circuits when exposed to ionizing radiation and assess their soft error rate (SER), it is fundamental to accurately model these diverse physical processes. In this chapter, we present a comprehensive modeling and analysis of the diffusion and collection mechanisms of radiation-induced charges through a semiconductor device. Analytical formulations of the collected charge, collection current, and collection velocity are developed. These equations are further employed to establish an analytical formulation of the soft error rate (SER), explaining its exponential dependence on the critical charge of the circuit. This formulation also links the SER to various physical and technological parameters, as well as to the characteristics of the radiation.
Keywords
- complementary metal-oxide semiconductor (CMOS)
- critical charge
- diffusion-collection
- metal-oxide-semiconductor field-effect transistor (MOSFET)
- radiation effects
- semiconductor
- single event effects (SEE)
- soft error rate (SER)
1. Introduction
The Earth is constantly exposed to a continuous bombardment of cosmic rays, also known as primary cosmic rays. These cosmic rays are a mixture of highly charged, energetic particles, consisting mainly of protons, helium nuclei, and heavy ions [1]. They come from various regions of deep space as well as from the Sun and arrive at Earth from all directions [2]. In the Earth’s upper atmosphere, the primary cosmic rays interact with atmospheric particles. The result is the formation of extended atmospheric showers, which produces secondary particles that reach sea level. At ground level, this produces a continuous flow of energy-distributed particles. Some of these particles (such as muons, neutrons, protons, and pions) interact with the materials of electronic devices and circuits and can cause undesirable effects on their operation or performances [3]. An important category of effects consists of single-event effects (SEEs), defined as any measurable or observable change in the state, operation, or performance of a microelectronic device, element, subsystem, or system (digital or analog), resulting from the impact of a single energetic particle [4].
An SEE is always triggered by the interaction of an incident particle with the target material. During the interaction, the incident particle necessarily transfers energy to the medium by electromagnetic or nuclear processes. As a result of these processes, some or all of the energy of the incident particles is released into the medium [5].
There are three main sequential steps involved in the production of SEE in microelectronic devices. The first step is the charge deposition as a result of the particle hitting the sensitive region of the device by one of two mechanisms: direct or indirect ionization [6]. Direct ionization involves charged particles interacting primarily with the electrons of the material’s atoms. A large number of excited energetic electrons (delta rays), usually energetic enough to ionize other atoms, are produced by the ionization mechanism. These electrons produce a cascade of secondary electrons, which lose their energy and create electron-hole pairs along the path of the particle. In this way, the deposited energy is essentially converted into electron-hole pairs. The energy required to create an electron-hole pair depends on the band gap of the material. In a silicon substrate, one electron-hole pair is created for every 3.6 eV of energy lost by the particle. Other particles (e.g., neutrons) do not ionize the matter they pass through and do not interact directly with the material. However, these particles can produce SEEs because they are likely to undergo nuclear reactions with atoms in the material. This mechanism is known as indirect ionization. The result of these nuclear reactions is the production of recoils and fragments that, like the charged particles, can deposit energy along their path by direct ionization [6].
The second step in the production of SEEs is the transport of the charge produced in the dense column of electron-hole pairs. Charges are transported by two main mechanisms: the drift of the charge in the regions subjected to an electric field and the ambipolar diffusion of the charge in the neutral zones. Finally, carriers generated in the electron-hole pair column can undergo an additional physical mechanism: carrier recombination with other mobile carriers in the lattice.
The third step is the collection of transported charges by elementary structures in the device (such as biased contacts or reverse-biased junctions). This charge collection process causes a parasitic transient current to be generated. This current is then injected into the node of the circuit affected by the particle and can cause disturbances in the circuit operation, such as SEEs.
Accurate modeling of these diverse physical processes is essential to predict the transient electrical behavior of complementary metal-oxide semiconductor (CMOS) devices and circuits when exposed to ionizing radiation and to assess their soft error rate (SER). The SER is defined as the probability that an ionizing particle will cause a transient error in the circuit that affects its operation, without causing permanent damage [3, 6]. In this chapter, we focus on the mechanisms of diffusion and collection of charges induced by energetic particles interacting with the materials of CMOS devices and circuits. A large amount of simulation and modeling studies have been carried out over the last 40 years in the field of radiation-induced diffusion-collection of charges in CMOS devices [7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21]. However, there are always challenges to be overcome concerning the calculation of the collection current and the evaluation of the circuit SER. This chapter presents a comprehensive modeling and analysis of radiation-induced charge diffusion and collection mechanisms through a semiconductor device. The aim of our development, based on the fundamental equations of diffusion-collection, is to reexamine some of the established results and to propose a new modeling of the collected charge, the collection current, and the collection velocity. Two different approaches are proposed, depending on the formalisms considered for the collection current: a pure diffusion current or a conduction current that includes a collection velocity. The equations obtained for the collected charge are then used to establish an analytical formulation of the soft error rate (SER), which explains its exponential dependence on the critical charge of the circuit. The chapter is structured as follows: In Section 2, we present the principle of the diffusion-collection model, and we recall the fundamental equations of the diffusion-collection mechanisms of the radiation-induced charges. These equations are the starting point of our modeling. Section 3 presents the diffusion-collection modeling based on the theory of diffusion. We start with the theoretical case of a point source and develop analytical expressions for the carrier density, diffusion current, and collected charge. The model accuracy is then verified using random-walk drift-diffusion (RWDD) numerical simulation. Next, we consider the more realistic case of charges that are generated along the path of an ionizing particle with a constant linear energy deposition, and we study the diffusion-collection of these charges. A linear distribution of point charges is used to emulate the particle track. An analytical formulation of the collected charge is proposed and validated through RWDD simulation. The collected charge model is then used to calculate the SER. In Section 4, the modeling of the collected charge and current is developed within a formalism that includes a collection velocity. After presenting the general case of a time- and position-dependent collection velocity, we focus on two particular cases: a time-independent collection velocity and a constant collection velocity. In the latter case, we show that the exponential form of the SER, which has long been established empirically in the literature, can be obtained analytically. It relates the SER to various physical and technological parameters and the characteristics of the radiation.
2. Principle of diffusion-collection modeling
As explained in the introduction, when a charged particle traverses through a semiconductor material, the energy it releases is transformed into electron-hole pairs. In the framework of the diffusion-collection model, these pairs are then organized into a series of localized point charges, such as discrete clusters of electrons and holes, concentrated within very tiny regions of the semiconductor material and distributed along the particle track (Figure 1). The number of pairs immediately after the energy deposition and pair formation is denoted as δn and δp for electrons and holes, respectively, with δn = δp = δn0.
The physical mechanism through which these excess charges (electrons and holes) are transported is ambipolar diffusion, particularly in the neutral zones of the device where there is no electric field. The model then assumes that a three-dimensional (3D) spherical diffusion law in the semiconductor domain governs the transport of these excess carriers. In the case of electrons, the time and space variation of the excess electron density
where
The solution of Eq. (1) gives the excess carrier density at the time
To maintain the analytical nature of the model, we neglect carrier recombination in Eq. (3). This is an important approximation, but it simplifies the calculations in the following to a considerable extent. The new excess carrier density equation is as follows:
As explained above, the charge transported by ambipolar diffusion is then collected by collecting structures of CMOS devices. The charge collected at point
3. Diffusion-collection modeling based on the theory of diffusion
3.1 Collected charge from a point source
We start with the theoretical case of a point source of radiation-induced charges located at
3.1.1 Diffusion current and collected charge modeling
The diffusion current density is given by the following:
The flow of charges transported by diffusion is collected at the surface of the semiconductor by a very small area contact centered at the
where
From the collected current, the charge collected at the contact at the
where Γ(a, x) is the upper incomplete gamma function [22]. As time tends toward infinity, the gamma function in Eq. (8) becomes equal to
The total charge collected at the
3.1.2 Verification by RWDD numerical simulation
We performed extensive numerical random walk drift-diffusion simulations to validate the diffusion current and collected charge models presented in Eqs. (7)–(9). These simulations included point sources of particles positioned at various distances from the collecting contact. We recall that the random-walk drift-diffusion (RWDD) model is a particle Monte Carlo technique. It effectively combines drift and diffusion processes using a random-walk algorithm that mimics the Brownian motion [23, 24, 25]. In the context of RWDD, the trajectory of an ionizing particle passing through the semiconductor region of devices is represented as a sequence of charge packets distributed along a linear segment. The length of this segment is equal to the range of the ionizing particle within the material. The precision of this charge discretization is guaranteed by adjusting the level of “granularity.” This can be fine-tuned by choosing the packet size, which in practice typically ranges from 1 to 100 elementary charges. The transport and the recombination of the electron and hole charge packets begin immediately after the particle passes through the device. Excess carriers undergo both diffusion and drift within the preexisting “background” electric field that existed before the particle impact in the equilibrium state. The drift-diffusion motion from time
where
An illustration of the charge transport simulation after the generation of electron-hole pairs in the particle Monte Carlo RWDD approach is shown in Figure 3. In this simulation, we consider a radiation-induced punctual charge
We use RWDD simulation to verify and validate the analytical modeling of the collected charge given by Eqs. (8) and (9). Figure 4 shows the time evolution of the charge collected by the contact given by Eq. (8) and by RWDD simulation with the parameters considered in Figure 3. We observe good agreement between the two curves shown in Figure 4. In addition, the maximum collected charge predicted by Eq. (9) closely matches that obtained in the RWDD simulation. This demonstrates the validity of Eqs. (8) and (9) in describing the time dependence of the collected charge. Figure 4 also shows the stochastic nature of the RWDD curve due to the small amount of charge collected, which is about a few tens of electrons at the end of the transient.
3.2 Collected charge from a particle track
We consider now the more realistic case of charges that are generated along the path of an ionizing particle with a constant linear energy deposition. We examine a simplified scenario, as shown in Figure 5, in which an ionizing particle strikes the semiconductor surface perpendicularly. It enters the semiconductor at location I, moves through the semiconductor over a path equal to its range, and stops at location S.
The particle loses the energy Δ
where
The particle track is treated as a set of point charges
where
The total collected charge at P point,
From (13), the total collected charge at P point is given by the following:
with
Eq. (14) has been validated by RWDD numerical simulation. We have considered an incident particle with a track of 1.5 μm hitting the semiconductor at a distance of 0.75 μm from the contact. A sequence of four different moments captured at
3.3 SER modeling
To determine the SER of the circuit, we use the notion of critical charge, which is a key parameter for assessing the susceptibility to radiation of the circuit. The critical charge
According to Eq. (16), a particle hitting the semiconductor at a distance
where
Usually, the critical charge
To express the solution of Eq. (16) analytically, we use a well-known approximation for the arctan function as follows:
We set
Then, Eq. (16) becomes.
where
The positive solution of the second-degree polynomial in Eq. (20) directly gives
The SER can be then calculated from Eqs. (17) and (22) as follows:
We have compared the SER given by Eq. (23) with the SER obtained as the exact solution of Eq. (16) solved numerically by the dichotomy method. Figure 8 shows the results of this comparison for three particle track lengths. In this figure, the SER is plotted as a function of critical charge for both calculation methods. A good agreement between the analytical model and the numerical solution is obtained, in particular for large values of the critical charge.
4. Diffusion-collection based on the collection velocity
As explained before, the general kinetic formulation of conduction is a second formalism that can be used to calculate the density of the current collected at the contact coming from the charges generated along the ionizing particle track. In this section, we will use this formalism to express the current density and the collected charge, as well as the SER of the circuit. This formalism introduces the notion of collecting velocity, which leads to several approximations related to its dependence on time and space, as we will see below.
4.1 Conduction current
In the general kinetic formulation of conduction, the current density is given by the following:
where
The current density, whether expressed in diffusion theory or in the general formulation of conduction theory, must be the same. Therefore, Eqs. (5) and (24) must be equal as follows:
which results in the following equation for the collection velocity:
From Eq. (27), vcol can be also rewritten as follows:
Eq. (28), which gives the general formula for the collection velocity, indicates that
It is obvious that integrating the collection current given by Eq. (29) over time from 0 to infinity will give the same collected charge as that given by Eq. (9). In this case, the calculation of the total charge and the SER will also be the same as that described in the previous section. We therefore obtain the same collected charge using both formalisms for the current density. This is, of course, an expected result. In fact, the interest in using this second formalism based on the collection velocity lies in the fact that several approximations can be made to this collection velocity to simplify the calculation of the collected charge and of the SER. In the following, we will focus on these approximations of the collection velocity.
4.2 Time-independent, position-dependent collection velocity
A first approximation used in many studies [10, 13, 14, 19] is to include a time-independent collection velocity in the expression (25) for the collection current as follows:
where
As time tends toward infinity, the gamma function in Eq. (31) becomes equal to
The amount of charge collected given by Eq. (32) must equal the amount of charge given by Eq. (9). This results in the following equation for the collection velocity:
As expected, this collection velocity is independent of time and depends only on the position
The maximum collected charge given by Eq. (32) becomes:
The total collected charge is obtained, similar to Eq. (13) and using (12), by integrating Eq. (35) along the entire particle track from 0 to
The total collected charge is then given by the following:
where
Next, we calculate the SER according to the relation (17) in the same way as in Section 3.3. The length
The SER can be then calculated from Eqs. (17) and (39) as follows:
The SER derived from Eq. (40) and that calculated by the numerical solving of Eq. (16) were compared in detail, as shown in Figure 9. Three values of particle track length were considered in this figure. A very good agreement between the analytical model and the numerical solution is observed over the whole range of critical charge values. This shows that assuming a time-independent collection velocity with a value averaged over the track length leads to an acceptable approximation of the SER compared to the exact numerical solution. The advantage of this assumption is that it simplifies the calculation of the SER and provides an analytical solution.
4.3 Constant collection velocity
A second approximation, more restrictive than the previous one, is to consider a constant collection velocity (a single numerical value that therefore does not depend on time or position) in the expression of the collected charge (32) [21]. We denote as vCOL this constant velocity. Eq. (32) becomes as follows:
The total collected charge is obtained by integrating Eq. (41) along the entire particle track from 0 to
where
The next step is to calculate
Finally, the SER can be calculated from Eqs. (17) and (44) as [21]:
As explained in [21], the expression of the SER (45) including a constant value for the collection velocity makes it possible to mathematically demonstrate the exponential dependence of the SER on the critical charge, proposed more than 20 years ago by Hazucha and Svensson in [28] as follows:
where
Assuming that
and then the SER is given in this case by the following:
Thus, we find an exponential dependence of the SER on the critical charge such as that of Eq. (44) that has been proposed empirically in the literature [28]. Eq. (49) clearly shows that this exponential dependence has a physical basis, being the result of the collection-diffusion mechanism of radiation-induced charges in the semiconductor.
In Figure 10, we compare the SER calculated by the exponential expression (49) with the SER obtained by the numerical solution of Eq. (16). We observe that the exponential expression matches the exact value of SER over a limited range of critical charge values (between 1.2 and 2.1 fC for
5. Conclusion
In this chapter, we presented a comprehensive modeling and analysis of the mechanisms of radiation-induced charge diffusion and collection through a semiconductor device. Our development proposes a new model of the collected charge, collection current, collection velocity, and soft error rate based on the fundamental diffusion-collection equations. The successive stages of the development are summarized in Figure 11, including the most important equations of the model. As shown in this figure, we have considered two different formalisms for the collection current: a pure diffusion current and a conduction current that includes a collection velocity. First, we considered the theoretical case of a point charge source and developed a model for the charge collected by a small contact. Second, this charge was used to model the charge collected in the more realistic case of charges generated along the track of an ionized particle crossing the device. The equations obtained for the collected charge were then used to establish an analytical formulation of the error rate as a function of the critical charge of the circuit. Two approximations can be used to simplify the calculations and obtain analytical expressions in the case of the formalism that includes a collection velocity. A first approximation is the use of a time-independent collection velocity and the integration of an average of this velocity along the particle track into the expression for the collected charge. Our results showed that this assumption leads to an acceptable approximation of the SER compared to the exact numerical solution. A second approximation is the use of a constant collection velocity, which results in an analytical expression of the SER with an exponential dependence on the critical charge. This type of dependence has been proposed empirically in previous works. Our model showed that there is a physical basis for this exponential expression, which results from the charge collection and diffusion mechanism of the radiation-induced charges in the semiconductor.
References
- 1.
Valkovic V. Radioactivity in the Environment. Amsterdam: Elsevier; 2000 - 2.
Dorman LI. Cosmic Rays in the Earth’s Atmosphere and Underground. Dordrecht: Kluwer Academic Publishers; 2004 - 3.
Autran JL, Munteanu D. Soft Errors: From Particles to Circuits. Boca Raton: CRC Press; 2015 - 4.
JEDEC Standard. Measurement and reporting of alpha particle and terrestrial cosmic ray-induced soft errors in semiconductor devices. N° JESD89B, Revision of JESD89A. 2021 - 5.
Leroy C, Rancoita PG. Principles of Radiation Interaction Matter and Detection. Singapore: World Scientific Publishing Co. Pte. Ltd; 2004 - 6.
Munteanu D, Autran JL. Modeling and simulation of single-event effects in digital devices and ICs. IEEE Transactions on Nuclear Science. 2008; 55 :1854-1878 - 7.
Kirkpatrick S. Modeling diffusion and collection of charge from ionizing radiation in silicon devices. IEEE Transactions on Electron Devices. 1979; 26 :1742-1753 - 8.
Messenger GC. Collection of charge on junction nodes from ion tracks. IEEE Transactions on Nuclear Science. 1982; 29 :2024-2031 - 9.
Edmonds L. Charged collected by diffusion from an ion track under mixed boundary conditions. IEEE Transactions on Nuclear Science. 1991; 38 :834-837 - 10.
Palau JM, Hubert G, Coulie K, Sagnes B, Calvet M-C, Fourtine S. Device simulation study of the SEU sensitivity of SRAMs to internal ion tracks generated by nuclear reactions. IEEE Transactions on Nuclear Science. 2001; 48 :225-231 - 11.
Palau JM, Wrobel R, Castellani-Coulie K, Calvet M-C, Dodd P, Sexton F. Monte Carlo exploration of neutron-induced SEU-sensitive volumes in SRAMs. IEEE Transactions on Nuclear Science. 2002; 49 :3075-3081 - 12.
Lambert D, Baggio J, Ferlet-Cavrois V, Flament O, Saigne F, Sagnes B, et al. Neutron-induced SEU in bulk SRAMs in terrestrial environment: Simulations and experiments. IEEE Transactions on Nuclear Science. 2004; 51 :3435-3441 - 13.
Merelle T, Chabane H, Palau JM, Castellani-Coulie K, Wrobel F, Saigne F, et al. Criterion for SEU occurrence in SRAM deduced from circuit and device simulations in case of neutron-induced SER. IEEE Transactions on Nuclear Science. 2005; 52 :1148-1155 - 14.
Wrobel F, Hubert G, Iacconi P. A semi-empirical approach for heavy ion SEU cross section calculations. IEEE Transactions on Nuclear Science. 2006; 53 :3271-3276 - 15.
Correas V, Saigne F, Sagnes B, Boch J, Gasiot G, Giot D, et al. Innovative simulations of heavy ion cross sections in 130 nm CMOS SRAM. IEEE Transactions on Nuclear Science. 2007; 54 :2413-2418 - 16.
Correas V, Saigne F, Sagnes B, Boch J, Gasiot G, Giot D, et al. Simulation tool for the prediction of heavy ion cross section of innovative 130-nm SRAMs. IEEE Transactions on Nuclear Science. 2008; 55 :2036-2041 - 17.
Artola L, Hubert G, Duzellier S, Bezerra F. Collected charge analysis for a new transient model by TCAD simulation in 90 nm technology. IEEE Transactions on Nuclear Science. 2010; 57 :1869-1875 - 18.
Uznanski S, Gasiot G, Roche P, Tavernier C, Autran JL. Single event upset and multiple cell upset modeling in commercial bulk 65-nm CMOS SRAMs and flip-flops. IEEE Transactions on Nuclear Science. 2010; 57 :1876-1883 - 19.
Artola L, Hubert G, Warren KM, Gaillardin M, Schrimpf RD, Reed RA, et al. SEU prediction from SET modeling using multi-node collection in bulk transistors and SRAMs down to the 65 nm technology node. IEEE Transactions on Nuclear Science. 2011; 58 :1338-1346 - 20.
Martinie S, Autran JL, Sauze S, Munteanu D, Uznanski S, Roche P, et al. Underground experiment and modeling of alpha emitters induced soft-error rate in CMOS 65 nm SRAM. IEEE Transactions on Nuclear Science. 2012; 59 :1048-1053 - 21.
Autran JL, Munteanu D. Physics-based analytical formulation of the soft error rate in CMOS circuits. IEEE Transactions on Nuclear Science. 2023; 70 :782-791 - 22.
Paris RB. Incomplete gamma and related functions. In: Olver FWJ, Lozier DW, Boisvert RF, Clark CW, editors. NIST Handbook of Mathematical Functions. Cambridge: Cambridge University Press; 2010. Chapter 8 - 23.
Autran JL, Glorieux M, Munteanu D, Clerc S, Gasiot G, Roche P. Particle Monte Carlo modeling of single-event transient current and charge collection in integrated circuits. Microelectronics Reliability. 2014; 54 :2278-2283 - 24.
Glorieux M, Autran JL, Munteanu D, Clerc S, Gasiot G, Roche P. Random-walk drift-diffusion charge-collection model for reverse-biased junctions embedded in circuits. IEEE Transactions on Nuclear Science. 2014; 61 :3527-3534 - 25.
Autran JL, Munteanu D, Moindjie S, Saad Saoud T, Malherbe V, Gasiot G, et al. Charge collection physical Modeling for soft error rate computational simulation in digital circuits. In: Sher Akbar N, Anwar Beg O, editors. Modeling and Simulation in Engineering Sciences. Vienna: London, UKIntechOpen; 2016 Chapter 6 - 26.
Plante I, Cucinotta FA. Monte-Carlo simulation of particle diffusion in various geometries and application to chemistry and biology. In: Chan VWK, editor. Theory and Applications of Monte Carlo Simulations. Vienna: London, UKIntechOpen; 2013 - 27.
Press WH, Teukolsky SA, Vetterling WT, Flannery BP, Metcalf M. Numerical Recipes in c: The Art of Scientific Computing. 2nd ed. Cambridge: Cambridge University Press; 1992 - 28.
Hazucha P, Svensson C. Impact of CMOS technology scaling on the atmospheric neutron soft error rate. IEEE Transactions on Nuclear Science. 2000; 47 :2586-2594