Characterization and Modeling of a Highly Reliable ONO Antifuse for High-Performance FPGA and PROM
Liu Guozhu1, Wu Jianwei1, Yu Zongguang1, Hong Genshen1, Zheng Ruocheng1, Liu Baiqin1, Wu Suzhen1, Du Tao2
1Key Laboratory of Radiation Hardened Integrated Circuit, China Electronics Technology Group Corporation No.58 Research Institute, Wuxi, China
2State Key Laboratory of Electronic Thin Films and Integrated Devices, University of Electronic Science and Technology of China, Chengdu, China
To cite this article:
Liu Guozhu, Wu Jianwei, Yu Zongguang, Hong Genshen, Zheng Ruocheng, Liu Baiqin, Wu Suzhen, Du Tao. Characterization and Modeling of a Highly Reliable ONO Antifuse for High-Performance FPGA and PROM. International Journal of Materials Science and Applications. Vol. 5, No. 3, 2016, pp. 169-177. doi: 10.11648/j.ijmsa.20160503.19
Received: June 25, 2016; Accepted: July 14, 2016; Published: July 19, 2016
Abstract: The characteristics and reliability of ONO (oxide-nitride-oxide) anti-fuse devices prepared basing on 0.6µm SOI CMOS process have been studied experimentally and theoretically. The intrinsic principles of ONO dielectric breakdown were investigated with tunneling and emission models. It has been found that the conduction mechanisms of ONO dielectrics under high electric field (>10MV/cm) mainly obeys Fowler-Nordheim (F-N) tunneling and Poole-Frenkel (P-F) emission and Ohmic transport models. Meanwhile, the nitrogen depth distribution and the composition of the ONO stack films have been accurately determined using SIMS and EDX, respectively. The results indicate that the nitrogen concentration of interface between tunneling oxide and N+ sub-silicon is higher than that of interface between top oxide and N+ poly-silicon, which can contribute to prove the difference of top and bottom electrode interface barriers according to energy band diagrams of ONO anti-fuse devices. Besides, it is also found that the average breakdown voltage of ONO anti-fuse arrays and that of distribution decrease with increasing the number of anti-fuse cells, and the result is attributed to traps density of various areas. Moreover, the programming resistance of ONO anti-fuse cells and programming circuits decreases with increasing programming current, and the programming resistance of ONO anti-fuse cells can reach less than 200 ohm/cont when programming current is above 5mA. And the life of unprogrammed ONO anti-fuse devices can reach more than 40 years under an electric stress of 5.5v at the temperature from 25°C to 125°C. So, it can be concluded that the characteristics and reliability of the proposed ONO anti-fuse elements are suitable for applications in FPGA and PROM.
Keywords: Antifuse, Oxide-Nitride-Oxide, Breakdown Voltage, Programming Resistance, SOI
Over the past years, due to the ONO anti-fuse devices exhibiting great difference in impedance before and after programming, the anti-fuse processing technology has attracted significant attention of IC designers and manufacturers, and is widely used in the area of nonvolatile memories or programmable array logic devices for radiation-hardened space environment, such as PROM, PAL, FPGA (field programmable gate arrays), and so on [1-6]. The composition, structure, electrical characteristics and reliability of ONO antifuses have been widely studied. The ONO dielectrics layer lies between a poly-silicon conductor and a heavily doped N+ diffusion region of the base silicon wafer. Prior to programming, the anti-fuse cell tends to exhibit very high resistance (typically >1MΩ), and the impedance may be reduced to below 1000Ω after programming under high voltage or current stress [1-4]. Under the programming voltage, the ONO sandwich melts and the base wafer grows an epitaxial "bump" into the poly-silicon in the shape of a dome. Growth of the bump shatters the ONO layer, allowing diffusion of the substrate N+ into the poly-silicon to form a low-resistance path. The bump’s size also makes the structure tolerant of current-induced stresses [1-4]. If anything, the high currents would increase the bump’s size, strengthening the connection. Meanwhile, silicon-on-insulator (SOI) devices have superior capability in good radiation hardness, no latch-up, and high device density [7,8]. Therefore, to realize high performance and reliability of radiation-hardened ONO anti-fuse devices for FPGA or PROM, the SOI CMOS processing technology can be applied in the ONO anti-fuse process.
In this work, the electrical properties and reliability of ONO anti-fuse devices were investigated experimentally and theoretically basing on 0.6µm SOI CMOS process. Meanwhile, the N/O/Si depth distributions and composition of the ONO stack films have been accurately determined using SIMS and EDX, respectively. Besides, in order to characterize the electrical properties and reliability of ONO anti-fuse devices, current-voltage (I-V) and time-dependent dielectric breakdown (TDDB) were used and performed.
2.1. Preparation of the ONO Anti-fuse Samples
In this paper, the ONO anti-fuse samples were prepared basing on 0.6µm SOI CMOS process (2Wells@2Polys@2Metals). The anti-fuse ONO dielectrics is formed as follows: At first, after the formation of CMOS on the SOITEC P<100> (top-Si/ bottom-oxide=1.5µm/3.0µm), the high density of N+ sub-silicon in p-well is formed by arsenic and phosphorus implantation (dose: E15) and annealing at 1050°C. Since, the tunneling oxide layer is very thin, the process of which is difficultly controlled through conventional methods. Meanwhile, in order to improve tunneling oxide/ n+ sub-Silicon interface properties, the oxynitridation method is taken into account. Hence, bottom oxide (i.e., tunneling oxide) is thermally gown on n+ sub-silicon by the low-pressure oxynitridation inmixture of N2O and N2 ambients (i.e., N2O:O2=1:5), in which the temperature and pressure are fixed at 820°C and 70 torr, respectively . And following the rich-nitrigen silicon nitride is immediately deposited by the LPCVD method. The top-oxide is formed by re-oxidation of SixNy at 950°C (i.e., wet oxide). After that, a polysilicon layer is deposited and doped by phosphorus diffusion at 900°C utilizing a POCL3 source. The ONO anti-fuse structure is illustrated as Fig. 1.
2.2. Measurement and Method
In this study, the cross-section of an ONO anti-fuse device and the composition of ONO dielectrics were measured by TEM and EDX, respectively, as shown in the fig. 2. The thickness of tunneling oxide and media SixNy and top oxide are 2.9nm/4.8nm/3.0nm, respectively. Meanwhile, the nitrogen depth distribution and the composition of the films have been accurately determined using SIMS, and the depth profile of N/O/Si is shown in the fig. 3. To characterize the electrical properties of ONO anti-fuse dielectrics, capacitance-voltage (C-V) and current-voltage (I-V) measurements were carried out on MOS capacitors (i.e., N+ polysilicon gate / dielectrics / N-type sub-silicon). Besides, in order to characterize the electrical properties and the reliability of ONO anti-fuse devices, I-V and time-dependent dielectric breakdown (TDDB) were used and performed. I-V and high frequency (1MHz) C-V measurements were performed by HP model 4200 semiconductor parameter analyzer and HP model 4284 precision LCR meter, respectively.
3. Results and Discussion
3.1. ONO Conduction Mechanism
Fig. 4 shows I-V characteristics for ONO anti-fuse devices under different bias, and the effective electric field can be calculated approximately by V/Tono. It can be observed that the slopes of the I-V curves for ONO anti-fuse devices are approximately similar in the high-field region (>7.5MV/cm) with various bias. Meanwhile, the maximum breakdown fields are also similar when N+ sub-silicon and N+ poly-silicon electrodes added by +/- bias or -/+ bias, respectively. But the leakage current of low electric field is different, when N+ sub-silicon as top polarity, the leakage current in the low electric field is higher than that of N+ poly-silicon, and the difference has two to three orders of magnitude. Besides, the difference of breakdown voltage is approximately 2.5v under top and bottom electrodes with the same bias. The results can be contributed to the effect of interface roughness  and barrier height between cathode and ONO.
To study the conduction mechanism of the carriers in the ONO dielectrics, the current density (J) versus effective electric field (Eeff) data were fitted by Fowler-Nordheim(F-N) tunneling, ohmic transport and Poole-Frenkel(P-F) emission models, as shown in the format (1)~(5). The F-N model is based on the hot electron tunneling through a triangle-shaped energy barrier formed in the dielectric layer due to the large applied electric field. The P-F model describes the field enhanced thermal emission process of trapped electrons into the conduction band of the dielectrics .
Poole-Frenkel emission model:
Fowler-Nordheim tunneling model:
Ohmic transport model:
J3=C3E exp(-Ea/kT) (5)
Where J is the current density, E is the applied electric field, q is the elementary charge, ΦB is the potential barrier (in zero applied electric field) that an electron must cross to move from one atom to another in the crystal, εo is the dynamic permittivity, d is the thickness of the insulator, K is Boltzmann's constant, Kd is dielectric constant, T is the temperature, C1 is a function of the density of dielectrics trapping centers, C2 is a function of the electron effective mass and the depth of the trap-potential well, C3 is a function of the density of dielectrics trapping centers and the density of conducting carriers.
The energy band diagrams for ONO anti-fuse devices with various bias are shown in fig. 5. According to previous reports [12, 13, 14], the ONO anti-fuse conduction mechanism can be dedicated as follows.
When N+ polysilicon or N+ sub-silicon is applied with a positive bias, as shown in Fig. 5(a) and (b), holes are injected into the SixNy film by direct tunneling through the reoxidized film (top-layer) or tunneling oxide (bottom-layer), and they are carried by the P-F mechanism in the SixNy film (middle-layer). The holes will accumulate at the interface of the middle-layer and the bottom oxide or the middle-layer and the top-layer, although partly holes are compensated for F-N tunneling electron. Thus, such holes lead to electric field enhancement of bottom oxide (B.O) and enhance the F-N electron conduction in B.O.
However, when N+ polysilicon or N+ sub-silicon electrode is added by a negative bias, as shown in Fig. 5(c) and (d), holes are injected into the valence band of the SixNy film by the modified F-N tunneling because of the thin bottom oxide film, and they are carried by the P-F mechanism. For hole conduction from the SixNy film to the electrode, two mechanisms are thought to occur: Tunneling from the trap site of the SixNy film to the electrode through the oxide and tunneling from the valence band to the electrode. The tunneling conduction from the trap site is thought to be dominant.
As shown in the fig. 3, the SIMS results clearly demonstrate that the wet re-oxidation of Si3N4 results in a high concentration of nitrogen near ONO stack layers, and the nitrogen concentration of top-oxide and middle-oxide and tunneling oxide approximately are 24.43%, 40.15%, 29.93%, respectively. According to Yu etal  results of the relation between " the electron or hole barrier height and dielectric gap energy" and "the N concentration of oxynitride films", the band gap energy of top-oxide and middle-oxide and tunneling oxide approximately are 6.75ev (ΔEv: 3.15ev; ΔEc: 2.45ev；ΔEsi: 1.12ev), 5.8ev、6.25ev (ΔEv: 2.95ev; ΔEc: 2.15ev；ΔEsi: 1.12ev), as shown in the table 1.
Meanwhile, the tunneling / injection properties of hot electrons (HEs) and hot holes (HHs) via the potential barrier formed by tunneling oxide or top-oxide of ONO antifuse devices are investigated. According to the principle of F-N tunneling model as shown in the format (3), the electron and hole interface barrier height of oxynitride / N+sub-silicon and oxynitride / N+ poly-silicon with various temperatures for ONO anti-fuse devices are calculated basing on the slope of J/E2 vs 1/E in different bias condition, and shown in the fig. 6 and table 1, and that of ΔEc and ΔEv is in agreement with previous reports  at temperature 25°C. It can be shown that the interface barrier height of the electron and hole linearly increases or decreases with temperature from 20°C to 180°C, and the ONO breakdown voltage under different bias is strongly depended on the interface electron barrier height as shown in the fig. 7. It is understood that the decrease of ΔEc exponentially enhances the tunneling probability of HEs, while it exponentially suppresses the tunneling probability of HHs. The positive charge(HHs) produced by injection is observed such that the cathode field for F-N current injection is enhanced and the anode field is decreased, depending on the location of these charge [10-18]. The trapped positive charges resulting from hole generation by impact ionization in the oxide lead to an increased local field and reduce barrier height of the cathode under high electric field. The P-F transport model suggests that electron trapping causes an increase in the electric field at the anode and eventually leads to breaking of Si-O bands [10-18].
Fig. 8(a) and (b) show the Poole-Frenkel(P-F) and the Fowler-Nordheim(F-N) plots of the measured current-voltage relations for the ONO antifuse devices under various bias. The plots of log (J/Eeff) versus Eeff1/2 and log (J/Eeff2) versus Eeff-1 approximately behave in accordance with linear fitting, and that of adjective coefficients are more than 0.99, and which indeed indicates the exactness of the model. So the ONO conduction mechanism obeys either a P-F emission model or a F-N tunneling model. Meanwhile, the dielectric constant Kd can be extracted from slop of log (J/Eeff2) versus Eeff-1 using format (2). And basing on comparing with measured Kd from C-V characteristics of ONO anti-fuse capacitors as shown in the fig. 9 and the table 2, it is found that the matching between measured and calculated Kd is very perfect, which also prove our previous conclusion. In addition, Fig. 10 shows log (J/Eeff) versus 1000/T plots for temperatures ranging from 25°C to 200°C. The temperature-dependent current at the higher electric field is mainly due to current components J1 and J3. According to the P-F emission model, current density J1 is due to field-enhanced thermal excitation of trapped electrons into the dielectric conduction band .
|ONO stacker||Extracted from N+ polyslicon as anode under positive bias||Extracted from N+ sub-silicon as anode under positive bias||C-V measured value|
|Dielectric constant (Kd)||5.39||5.59||5.34|
3.2. ONO Characteristics
3.2.1. Current-Voltage Characteristics
Fig. 11 shows excellent breakdown characteristics and distribution of ONO anti-fuse arrays containing anti-fuse cells from 1 to 1K. The ONO anti-fuse cells perform low leakage current under a 5.5v electric stress. Meanwhile, it is found that the average breakdown voltage of ONO anti-fuse arrays decreases with increasing the number of anti-fuse cells, and the distribution of breakdown voltage also decreases with increasing the number of anti-fuse cells by means of the uniformity and cumulative failure of BV, namely, the uniformity is defined as the std divided by the average of BV. The uniformity of arrays from 1 to 1K decreases from 1.83% down to 0.75%. The result is attributed to traps density of various areas.
3.2.2. Programming Characteristics
Fig. 12 shows ONO anti-fuse programming resistance with programming time under various programming current. When programming current reaches above 0.75mA, the resistance of ONO anti-fuse cells becomes constant without changing under the increase of programming time, and that of characteristics further illustrates that the ONO anti-fuse cells have extremely short programming time (i.e., Tprogram=307μs). Meanwhile, in order to study the relation between programming resistance with programming current of the ONO anti-fuse cells, we choose two testing structures: an ONO anti-fuse cell and two categories programming circuits, which contains an MOS transistor and an ONO anti-fuse cell (i.e., AF1+MOS) or an MOS transistor and two anti-fuse cells (i.e., AF1+AF2+MOS). The programming resistance of ONO anti-fuse cells and programming circuits decreases with increasing programming current, and the programming resistance of ONO anti-fuse cells can reach less than 200 ohm/cont when programming current is above 2mA, and that of programming characteristics is shown in the fig. 13. Besides, to further illustrate the relation of programming current and resistant, the numerical analysis for the experience date is carried out as following format (6), where the factor A and B can be extracted as 948.19 and 0.7015 basing on a power fitting model, respectively.
3.3. Time Dependent Dielectric Breakdown (TDDB)
According to ohmic transport model and I-V characteristics of ONO anti-fuse cells, the thermal active energy (Ea) of ONO dielectrics is extracted from the slope of a log (J/Eeff) versus T-1 under high effective electric field (Eeff) using the format (2) and (5), and the Eeff is choose basing on electrode bias from 8MV/cm to 11MV/cm for N+ poly-silicon as anode and from 10.5MV/cm to 13MV/cm for N+ sub-silicon as anode, respectively. Meanwhile, it is also found that Ea versus Eeff forms a linear equation as shown in fig. 14, and the thermal active energy (Ea) of ONO dielectrics under a 5.5v electric stress condition are calculated as 0.343 and 0.349 for N+ poly-silicon and N+ sub-silicon as anode, respectively.
Besides, the results indeed indicate that the ONO antifuse devices conduction mechanism mainly obeys a P-F emission model or a F-N tunneling model or ohmic transport model. So the TDDB of ONO dielectrics should mainly be dependent on E model basing on Mc Pherson and Baglee18,19 thermochemical foundation, and the E model is expressed as an equation(7).
Ea=-0.04633Eeff + 0.57493 (8)
Where t50 is the median time to breakdown, A and B are constant coefficient and field acceleration parameter, respectively. According to slops of log (t50) versus Eeff under different temperature basing on the TDDB of ONO dielectrics shown in the fig. 15, the factor A and B can be extracted by E model (i.e., the value of A and B are about 4.38E14 and 3.3455, respectively). So, it can be calculated that the life of an ONO anti-fuse cell (0.6µm*0.6µm) under an electric stress of 5.5v at the temperature from 25°C to 125°C is more than 40 years according to the format (7) and (8).
Highly reliable ONO anti-fuse devices(0.6μm*0.6μm) with excellent electrical characteristics, such as low programming resistance, low programming time, and good BV uniformity, can be successfully prepared basing on 0.6µm SOI CMOS process. The intrinsic principles of ONO dielectric breakdown were investigated experimentally and theoretically with tunneling and emission models. Firstly, it has been found that the conduction mechanisms of ONO dielectrics under high electric field (>10MV/cm) mainly obeys Fowler-Nordheim (F-N) tunneling and Poole-frenkel (P-F) emission and Ohmic transport models. Meanwhile, according to the depth profile of N/O/Si for an ONO antifuse using SIMS, the difference of top and bottom electrode interface barriers is due to the difference of nitrogen concentration at interface for "tunneling oxide /N+ sub-silicon" and "top oxide / N+ poly-silicon". Besides, the average breakdown voltage of ONO anti-fuse arrays and that of distribution decrease with increasing the number of anti-fuse cells, and the uniformity of arrays from 1 to 1K decreases from 1.83% down to 0.75%. In addition, the programming resistance of ONO anti-fuse cells and programming circuits decreases with increasing programming current, and the programming resistance of ONO anti-fuse cells can reach less than 200 ohm/cont when programming current is above 5mA, and the programming time Tpp is below 3.07E-4s. Finally, it can be predicated that the life of unprogrammed ONO anti-fuse devices can reach more than 40 years under an electric stress (5.5v) at the temperature from 25°C to 125°C by means of TDDB. So, all these results strongly indicate that the characteristics and reliability of the proposed ONO anti-fuse elements are suitable for applications in nonvolatile memory devices, such as PROM, FPGA and so on.