Irradiated MOS Devices - IEEE Xplore

12 downloads 0 Views 653KB Size Report
Abstract. In this work we have investigated how gamma irradiation affects the tunneling conduction mechanism of a 20 nm thick oxide in MOS capacitors.
IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 45, NO. 3, JUNE 1998

1390

Modifications of Fowler-Nordheim Injection Characteristics in y Irradiated MOS Devices A. Scarpal, A Paccagnella', F. Montera', A. Candelori', G. Glxbaudo2, G. Pananakakis2, G. Ghdini3 and P. G. Fuochi4 I

Dipartimento di Elettronica e Informatica, Universita di Padova, via Gradenigo 6/a, 3 5 13 1 Padova, Italy LPCSENSERG, BP 257,38016 Grenoble, France 3 SGS-THOMSON Microelectronics, Central R & D, via C. Olivetti 2, 20041 Agrate Brianza, Italy 4 CNR-FRAE, via Gobetti 101, 40 129 Bologna, Italy

Abstract In this work we have investigated how gamma irradiation affects the tunneling conduction mechanism of a 20 nm thick oxide in MOS capacitors. The radiation induced positive charge is rapidly compensated by the injected electrons, and does not impact the gate current under positive injection after the first current-voltage measurement. Only a transient stress induced leakage current at low gate bias is observed. Instead, a radiation induced negative charge has been observed near the polysilicon gate,>which enhances the gate voltage needed for Fowler-Nordheim conduction at negative gate bias. No time decay d t k i s c h a y c has been observed. Such charges slightly iL?ui.iilry CL: kapping kinetics cf iicge!+e charge during subsequent electrical stresses performed at constant current condition.

I. INTRODUCTION effects on MOS devices have

Radiation attracted a large attention during the last decades, and a variety of studies has been performed by many authors. A few aspects of this subject have not been considered in detail: among them, the modification of the conduction characteristics across the irradiated oxide, which has attracted only a modest attention from the scientific community. This problem could be considered not important for thick oxides, but now MOS with oxide thickness under 10 nm are currently produced, and even more in the future the microelectronic industry will rely on thin oxide devices. In thin oxides, low gate voltages can easily reach the threshold oxide field for tunneling injection, and deep submicron technologies will face the problem of a gate contact with non-zero conductance. Tunneling injection is also positively exploited for writeierane operations of non volatile memories. The first aim of this work has been the study of the effects of gamma exposure on the current conduction characteristics of MOS capacitors with 20 nm oxide thickness. This oxide is not exactly ultra-thin, but it gives the advantage that standard experimental techniques based on the analysis of the Fowler-Nordheim tunneling (such as the DiMaria technique [ 11) can be successfully applied in this case to monitor part of the oxide charge distribution. In fact, by decreasing the oxide thickness, the quantitative results which can be obtained through these techniques, become less and

less meaningful, since no information can be collected on the charge accumulated inside oxide layers about 3 nm thick at both interfaces. This work will show how such experimental techniques can be applied for the study of the oxide trapping characteristics which cannot be investigated by using the current C-V and Id-V, MOSFET techniques. Tunnel injection has been also largely used to monitor the MOS reliability, through many different approaches (stresses at constant current, constant voltage, ramped voltage, etc.). The study of the evolution of trapped charge is a direct way of evaluating the reliability of the oxide. In fact, charge trapping can limit the device life time long before final breakdown. A +" J l l r l ~ ~f large interest, aiid still partially unexpiored [2j, i s representea oy tne evaiuatior, of oxide degradation, due to electrical stress following radiation exposure, These studies can be important for commercial device applications in radioactive environments. In fact, MOS devices degrade under operating voltages and currents, and this degradation could be affected by the further stress induced by ionizing radiation. In the second part of the work, some results from accelerated stresses on fresh and irradiated devices will be illustrated. - 7

11. EXPERIMENTALS A. Devices The devices used throughout this work are square MOS capacitors on p-Si, with gate area A = lo-* cm2 and oxide thickness to, = 20 nm. The MOS active area is surrounded by a N+ ring providing the minority carriers needed during electron injection from the substrate to the gate. In all devices the gate metal is a highly doped N+ polysilicon layer, while oxide was grown by a wet oxidation process. All devices have been fabricated by SGS-THOMSON Microelecaonics (Agrate Brianza, Italy).

B. Stresses In this work we have considered both radiation and electrical stresses on MOS devices. Radiation stresses have been performed at room temperature by using Co60 gamma rays from a Nordion Gammacell. The dose rate was 3.9 krad(Si)/min, up to final doses of 1 and 5.3 Mrad(Si). During and after irradiation,

0018-9499/98$10.00

0 1998 IEEE

1391

devicjs have been kept biased at gate-to-substrate bias values Vblas.t 3 V, with the N+ ring connected to the substrate contadt. Irradiated devices have been measured for the first time a out three hours after the end of the y exposure. El ctrical stresses have been performed by using the Const nt Current Stress (CCS) method. Electrons were always injected fiom the substrate (positive injection, i.e., V, > 0 during the stress) into the oxide, through a Fowler-Nordheim tunneling injection mechanism, up to a cumulative charge of N,,= I 10’’ electrons/cm2. The gate current density has been kept $xed at Jccs = 10 mA/cm2 during the stress, with a corresbonding V, a 17 V. These conditions ensure a constant oxide ,field at the cathode, as deduced from the expression of the Folwler-Nordheim (FN) current I, for electrons tunneling across1a triangular barrier [ 3 ] :

P

wherel F, is the oxide cathodic field, and A and B are two coeffidients depending on the oxide characteristics and independent of F,.

C. Measurements

-

X AV; _ --cox.-

to,

AQox

Capacitance versus gate voltage (C-V,) has been measured at a frequency f = 400 kHz, in order to overcome problems of the experimental set-up at higher frequencies. Measurements have been performed by using a HP 4284A LCR meter, but the voltage source was supplied by a HP4 140B instrument. All measurements have been performed on devices microbonded to DIL packages.

D. Test Procedure The experimental procedure is illustrated in the flow-chart in Fig. 1. All devices have been subjected to a preliminary electrical characterization before stresses, Le., Tests A, consisting of 1,-V, and C-V, measurements. Tests A have been repeated at different times after gamma irradiation, to study the annealing behavior of the oxide charge at room temperature. Different devices have been used for different storage times. The last Tests A have been followed by a CCS, which has been periodically stopped in order to measure the Fowler-Nordheim injection curves (Tests B). For comparison, non-irradiated devices have been subjected to the tests B as weii.

W t have performed different electrical measurements on MOS hevices befcie and a k r radiation aud dectricai stresses. Gate durrent versus gate voltage (Is-VJ has been measured for both dositive and negative V, sweeps at a ramp rate dlV,j/dt = 39 mv/sec, by using a HP4155A parameter analyzer. In particdlar, we have recorded the gate voltages needed to inject a gatel current density J,, = 0.1 mA/cm2, for both positive and negative injection. That is, when J,,, = +0.1 mA/cm*, electrdn are injected from the substrate and a gate voltage Vg+ is measured; correspondingly, when J,, = -0.1 mA/cmZ, electrcm are injected from the gate and Vg- is needed. The I shifts of these Fowler-Nordheim voltages, AV; and AV; respectively, after a radiatiordelectrical stress are related to the net charge induced in the oxide hy the stress itself, AQOx,and to the ~barycentreof the charge density distribution (centroid), x , as [I]: Figure 1. Experimental flow-chart.

111. RESULTS A. Eflects of Radiation Stresses

where C,, is the oxide capacitance and to, is the oxide thickness. The centroid position has been referred to the Si/SiQ, interface, that is, ?= 0 when the effective charge is at this interface. Equations (1) and (2) can be easily rearranged to cakulate AQoxand :

x

Figure 2 shows three consecutive positive Ig-Vg measurements performed 3 hr after the 5.3 Mrad(Si) irradiation: they are the first electrical measurements taken on the device after the gamma exposure. The first curve is much higher than the second one, which in turn overlaps quite well the thud curve. Even after the I Mrad(Si) y dose we observed the same behavior (not shown here), but with a smaller

1392

difference between the first two curves. A similar measurement instability appears on the irradiated devices every time the gate voltage polarity is switched (positiveinegative, or the opposite as well), but its magnitude is much lower than that of Fig. 2 .

indications on the charge stability under gate current injection. We'll codsider first the results for positive charge. I

l,lo.~o

1

1.10-l1 1 .

1

I

I

-

1

/

fresh fresh o

I

I _ _ 1_ - - -W

I ~

1.10 " O

5

1 10.120

v, n/, (a)

1.1 0." I

20

15

10

,,F

I.,"

0

5

10

v, (V)

15

20

,

t-

Figure 2. Three consecutive positive injection measurements performed on a MOS irradiated device (5.3 Mrad(Si)). Figures 3.a and 3.b show the stabilized curves (third or fourth in a series of measurements) for positive and negative 1,-V, measurements, respectively, on devices irradiated at a dose of 1 Mrad(Si). Modest radiation induced effects appear at low voltages, with some deformation of the low field leakage current. At h&igate valtages in the E P J LIlJeCKiGI regin?, I?C difference is observed in the posiave curve benveen irradiated and non-irradiated capacitors. On the other hand, a shift of the negative curves is observed toward more negative gate voltage values in the irradiated devices, without any noticeable difference for different storage times. Similar plots can be drawn for 5.3 Mrad(Si) irradiated devices, but with a larger shift of the negative current curves after irradiation. By using the FN curves of Fig. 3, and equations (3) and (4), we have evaluated the negative trapped charge density AQox = - 0 . 6 1~0-7 C/cm2 after 1 Mrad(Si), and AQ,, = - 1.1x 10-7 C/cm2 after 5.3 Mrad(Si), respectively. The charge centroid appears in both case close to the gate/oxide interface. in order io funher investigate the transient behavior of the 1,-V, positive curves, we have drawn a C-V, curve as the first measurement of a MOS capacitor after the radiation exposure. Then, another C-V, followed after 1,-V, measurements, as shown in Fig. 4. The gamma irradiation mduces a large negative shift, indicating a very large oxide positive charge, equivalent to a 5 ~ 1 0 ' ~ c mstate - ~ density, if this charge is assumed at the onide/silicon interface. After the Ig-Vg curves,

a positive threshold voltage shift appears, indicating the neutralization of part of this positive charge, with a residual 4 . 4 1~0 " ~ m -state ~ density left. If other C-V, measurements are performed after 24 hr and 240 hr,no further modifications of the curves are observed. Our results give evidence for radiation induced trapping of both positive and negative charge in our MOS capacitors, which are not unexpected, to say the least. Moreover, and this is the original part of this work, the results give also some

1.10.~

4

1.10

Figure 3. Is-Vg curves measured before (thick lines) and after 1 Mrad(Si) (thin lines: 3 hr annealing; dashed lines: 24 hr annealing; dotted lines: 240 hr annealing) irradiation, for (a) positive and (b) negative gate voltages

0.9 0.8

,

07

'l

before !g vg

0.5 0.4

0.3' -7

'

'

-5

'

'

'

-3

v,

'

-1

'

1

(v)

Figure 4. Capacitance-voltage measurements carried on fresh (thick line) and 5.3 Mrad(Si) irradiated devices (thin lines), before (continuos line) and after (dashed line) &-Vg measurements. The anomalous 1,-V, first curve after irradiation (see Fig. 2) can been attributed to two different mechanisms: the effect of the radiation generated positive charge and its neutralization, and the chargingldischarging of electron traps, as discussed in the following.

1393

In the first case, radiation stress induces hole trapping in the oxide near the Si/Si02 interface, as confirmed by the C-V, measurements. The effect of positive charge trapped near the cathode is to enhance the electron tunneling probability by reducing the area of the triangular barrier seen by the electrons [4]. Consequently, the Fowler-Nordheim tunneling current will increase just after irradiation, as observed during the first I,-Vg rneasurement following the stress. Such enhancement of the gate current could be read as another type of Radiation Induced Leakage Current (RILC), different from that recently observed by our group in irradiated ultra-thin (4.4 nm) oxides [ 5 ] . This current on the 20 nm oxides is not stable over subsequent measurements as in case of the 4.4 nm oxides, and the conduction mechanism appears substantially different from the trap-assisted one generally (if not universally) claimed as responsible m case of low field leakage current. If trapped holes can enhance the tunneling probability, still they can easily be neutralized and annealed by the electrons of the FN current [6], as indicated also by the C-V, measurements of Fig. 4. Such recombination processes can only take place in the oxide at a distance from the cathode (the Si substrate) larger than the FN tunneling distance, which is approximately 3 nm. Within the tunneling distance from the cathode, tunneling of electrons not participating to the FN conduction should be considered for the hole recombination, or for negative charging of border states inside the oxide. The idet.:: zf3 metastable electron trap asscciateri ?:.lba bole trap was f r s t discussed by Lelis et al. [7, 81 and then extended by Walters and Reisman [9]. Therefore, during the first current measurement, the trappeil positive charge is progressively reduced, hence decreasing the tunnel probability. At high applied voltages, that is, over 15 V in Fig. 2, the effect of the residual trapped positive charge has become negligible. This does not mean that all trapped holes have been compensated or recombined, as shown by the residual shift of the C-V, curves in Fig. 4. Simply, the amount and the position of the positive charge, likely close to the Si surface, can affect much more the threshold voltage than the current injection characteristics. Incidentally (and qualitatively), we would like to note here that a positive charge layer at the Si surface would not affect at all the F-N injection from Si [ 101. Quantitatively: the impact of a charge layer inside the tunneling distance on the F-N curve requires the calculation of the tunneling probability across a barrier which is no more triangular. This specific (and massive) problem has not been addressed here. Let's come now to the chargingdischarging of electron traps in the oxide, the second contribution to the anomalously high gate current. When I,-V, measurements are performed for the first time after the irradiation, electronic traps far from the cathode at least one F-N tunneling distance can capture an F-N electron injected from the Si substrate, while electrons not contributing to the F-N conduction can tunnel toward positive sites: in both cases, a net excess current transient is induced. A similar displacement current, but much less intense, is observed also during the first measurement after the gate

voltage polarity is switched, in agreement with previous data on electrically stressed devices [ 11, 121. In fact, a fraction of the trapped electrons will be emitted from the traps when the gate voltage is brought again to zero, but some traps with energy level under the cathode Fermi level will keep the trapped carrier, emitting it only when the gate polarity is reversed. We consider shortly the negative trapped charge. Negative values of AV; (see Fig.3.b) indicate that gamma irradiation have induced a negative charge (see eq. (I)). This is located close to the polysilicon gate/Si02 interface, as can be inferred by observing that AV; shifts are practically zero (see eq. (2)). The build-up of net negative charge in the oxide bulk after ionizing radiation stress is not often reported in the literature [13]: being far from the Si/Si02 interface, it can hardly be detected by measuring the flat-band or the threshold voltage shifts (see eq. 5), which are two common parameters adopted to study the oxide charge. When electron-hole pairs are generated in the oxide by ionizing radiation, negative carriers are rapidly swept toward the Si/gate interface, due to the positive applied electric field (V, = 3 V in our experiments). Despite their high mobility, same electrons are trapped in the background oxide sites, located near the Si/gate interface.

B. Constant Current Stress Results Irradiaiad and non-L~sdiatcddevices have btcr, rl Ihittec! to CCS tests. By using equations (3) and (4), we have calculated the oxide charge build-up and the corresponding charge centroid after each stress step, as shown in Fig. 5 and in Fig, 6, respectively.

a P

U

0.5

1.10"

1.1Ol8

1.10"

l.lom

Nini (electrons/cm2)

Figure 5. Oxide trapped charge due to CCS applied to fresh (line) and irradiated devices (filled symbols: 1 Mrad(Si); open symbols: 5.3 Mrad(Si)). Annealing time has been indicated as well (circles: 3 hr annealing; squares: 24 hr annealing; triangles: 240 hr annealing). We want to remark that AQOxindicates the charge induced only by constant current stress, regardless the MOS were fresh or irradiated before CCS. During such positive stresses (i. e. electrons injected from the substrate), a negative charge builds up in the oxide (see Fig. 5 ) , which increases with the electron fluence. Correspondingly, its centroid moves from the cathodic interface (i. e., the gate/SiO, interface) at low

1394

electron fluency toward the middle point of the oxide. This behavior is typical for unirradiated (fresh) devices, subjected to constant current or voltage stresses [14-161. However, it was never investigated on irradiated devices. Irradiated devices show slightly lower values of the stress induced negative charge, with respect to fresh ones. This difference can be attributed to the fact that, during irradiation, part of the electron background traps have already been filled by radiation induced electrons, as discussed in the previous section. When CCS is performed, these traps are not available to capture the injected carriers and, therefore, the differential trapped charge is lower. Practically, no difference is observed in the centroid position between irradiated and fresh device centroids, with the exception of the device stored for 240 hr. before CCS. A long time annealing could produce redistribution and modifications of the radiation induced trapped charge and of the radiation induced defects, due to the applied oxide field during the storage time. The centroid evolution for long time annealed device may be attributed to a different distribution of filled traps, but it eventually follows the other curves, as the stress level increases.

I:

~

J

non-irradiated

valid Lor be;:: ~ ~ z ! f - t .322 ~ ? nxi-iiabiaied devices. The device lifetime can be immediately deduced, as long as it is known the maximum oxide charge which permits the device correct operation.

N

020.5 v)

measurement precision, due to the low AQ,, values. At higher electron fluency, all the devices (irradiated and unirradiated) follow the same power law, with an exponent close to 0.7. Therefore, trapping rate is not noticeably affected by an irradiation stress, and the kinetics depends only on CCS. Hence, some considerations can be pointed out at least for high electron fluency: 0 a radiation stress does not induce any large modification of the electron trapping properties of the oxide as deduced from CCS tests, but only some background electron sites are filled; 0 radiation induced positive charge does not affect oxide electrical reliability. In fact, as electrons are injected throughout the SiO,, hoies are rapidly neutralized, while interface positive charge does not noticeably affect tunnel probability; 0 the difference of the negative trapped charge between irradiated and non-irradiated devices is an offset, which can be considered constant in a first order approximation. Therefore, the power law expressed by ( 5 ) can be actually considered an universal law. By cntegrating equation (5), the negative oxide charge can be expressed as:

15.3 Mrad(Si) Cn I

1 a

24 hr

1 1,

._

In .-

d

I

1101~

I

6

I

non-irradiated (line)

0

1.1Ol8

1.10’~

1.1oZ0

Nini (electronslcm?

-

6

1.10

2-

Figure 6. Charge distribution centroid due to CCS applied to fresh and irradiated devices

In previous works it has been shown that the kinetics of negative charge build-up follows an universal power law [4, 14, 161. This law can be expressed as: (5) where: Q,, is the CCS induced oxide charge, N,, the injected electron fluence, K a constant dependent on the oxide technology and v a positive exponent, quite independent of the oxide technology, with a value in the range 0.65 - 0.7 [4]. Fig. 7 shows the oxide charge build-up kinetics for fresh and irradiated devices. The behavior at low injected fluency shows again some differences between the fresh and irradiated devices, being lower the trapping rate in the latter ones, owing to the negative trapped charge pre-existing to the CCS test. However, in this regime results could also be affected by

2

- 1.10.~ 1.10

1 10”

1.10lS

110”

1 lo2p

N,,,(electrons/cmz) Figure 7 . Charge build-up kinetics due to CCS applied to fresh and irradiated devices.

IV. CONCLUSIONS The FN conduction characteristics of irradiated oxides 20 nm thick are modified by gamma irradiation. In particular, we have observed a transient radiation induced leakage current, deriving from positive charge trapped in the oxide. As injection proceeds, this positive charge is compensated and the anomalous RILC disappears, in a way similar to that reported for transient SILC after electrical stresses. On the other side, the negative trapped charge due to radiation leads to a stable shift of the FN curves under negative injection. Hence,

1395

standard measurements based on the FN tunneling can give quantitative evaluation at least of the oxide negative trapped charge, which is not commonly found on irradiated MOS devices. Non-perturbing methods aiming to measure the positive trapped charge could also be improved, for instance by performing punctual measurements of the FN voltage only at a given current density, with a minimum injected charge into the oxide. By considering the negative oxide charge, we have also studied the synergetic effects of gamma and constant current stress on the trapping characteristics of the oxide. In a fxst order analysis, the two stresses add the corresponding effects, with a reduction of the CCS negative charge due to the presence of the pre-existing negative charge deriving from the gamma exposure. The radiation stress appears ineffective in modifying the MOS device lifetime, in agreement with previous results obtained in different experimental conditions [2].

V. REFERENCES D. DiMaria, "Determination of Insulator Bulk Trapped Charge Densities and Centroids from Photocurrent-Voltage Characteristics of MOS structures", J. Appl. Phys., Vol. 47, NO. 9, pp. 4073-4077, 1976 A. .4sssime, G. J. Sarrabayrouse, G. Salace and C. Petit, "IICgradation Induire par I ' ii tadiatioii Gamma dans les Couches d'oxyile ile ~iiiciilm Ultra-tvlinces", ?roc. sf 3rd RADECS, pp. 80-84, Arcachon (France), September 1995 M. Lenzlinger and E. H. Snow, "Fowler-Nordheim Tunneling on Thermally Grown SiO,", J. Appl. Phys., vol. 40, No. I, PI). 278-284, 1969

C. Papadas, G. Ghibaudo, F. Pio, G. Pananakakis, P. Mortini and C. Riva, "On the Charge Build-Up Mechanisms in Gate Dielectrics", Solid-state Electron., vol. 37, No. 3, pp. 495-505, 1994 A. Scarpa, A. Paccagnella, F. Montera, G. Ghibaudo, G. Pananakakis, G. Ghidini and P. G. Fuochi, "Ionizing Radiation Induced Leakage Current on Ultra-Thin Gate Oxides", submitied to iEEE

Trans. on Nuci. Sei.

[6] K. Kobayashi, A. Teramoto, Y. Matsui, M. Hirayama and A. Yasuoka, "Electron Traps and Excess Current Induced by HotHole Injection into Thin SiO, Films", J. Electrochemic. SOC., V O ~ .143, NO. 10, pp. 3377-3383, 1996 [7] A. J. Lelis et al., "Reversibility of trapped hole annealing IEEE Trans. on iVucl. Sci., vol. 35, No. 6, pp. 1186-1191, 1988.

'I,

[SI A. J. Lelis et ai., "The nature of the trapped hole annealing process", IEEE Trans. on Nucl. Sci., vol. 36, No. 6, pp. 18081815, 1989 [9] M. Walters and A. Reisman, "Radiation-Induced Neutral Electron Trap Generation in Electrically Biased Insulated Gate Field Effect Transistor Gate Insulators", J. Electrochem. Soc., VOI.138, NO.9, pp. 2756-2762, 1991 [IO] B. Balland and G. Barbottin, "Trapping and Detrapping of Carriers injected into SiO,", Instabilities in Silicon Devices, G. Barbottin and A. Vapaille (Eds.), p. 61, New York, 1989 [ 1 11 R. S. Scott and D. J. Dumin, "The Charging and Discharging of

High-Voltage Stress-Generated traps in Thin Silicon Oxide", IEEE Trans. on Electron Devices, vol. 43, No. 1, pp. 130-136, 1996 [12] K. Sakakibara, N. Ajika, M. Hatanaka and H. Miyoshi, "A Quantitative Analysis of Stress Induced Excess Current (SIEC) in SiO, Films", Proc. of 34Ih IRPS., pp. 100-107, Dallas (U.S.A.), April 1996 [I31 P. Paillet, D. HervC, 5. L. Leray and R. A. B. Devine, "Evidence of Negative Charge Trapping in High Temperature Annealed rd : i " m ! Oxide"; Proc. of 2 RJDECS, pp. 140-145, Saint-kitialo (France;, S i P L i i k i Sf5 [14] A. Scarpa, G. Ghibaudo, G. Ghidini, G. Pananakakis and A. Paccagnela, "Stress induced degradation features of very thin gate oxides", Proc. of gh ESPRIT Workshop on Dielectrics in Microelectronics, Venice (Italy), November 1996 1151 E. Vincent, C. Papadas, C. Riva and G. Ghibaudo, "On The Charge Build-Up Mechanisms in Very Thin Insulator Layers", Proc. of ESSDERC '96,.pp. 495-498, Bologna (Italy), September 1996 [16] C. Papadas, G. Ghibaudo, C. Monserie and G. Panananakakais, "Reliability Issues of Silicon-Dioxide Structures - Application to FLOTOX EEPROM Cells", Microelectron. Reliub., Vol. 33, NO. 11/12 pp. 1867-1908, 1993