We have demonstrated that it is possible to reproducibly quantify hydrogen concentration in the SiN layer of a SiO2/SiN/SiO2 (ONO) stack structure using ultraviolet laser-assisted atom probe tomography (APT). The concentration of hydrogen atoms detected using APT increased gradually during the analysis, which could be explained by the effect of hydrogen adsorption from residual gas in the vacuum chamber onto the specimen surface. The amount of adsorbed hydrogen in the SiN layer was estimated by analyzing another SiN layer with an extremely low hydrogen concentration (<0.2 at. %). Thus, by subtracting the concentration of adsorbed hydrogen, the actual hydrogen concentration in the SiN layer was quantified as approximately 1.0 at. %. This result was consistent with that obtained by elastic recoil detection analysis (ERDA), which confirmed the accuracy of the APT quantification. The present results indicate that APT enables the imaging of the three-dimensional distribution of hydrogen atoms in actual devices at a sub-nanometer scale.

Demands for the reliability of semiconductor devices, especially in the automotive industry, are continuously increasing.1 Hydrogen atoms lead to severe degradation of reliability, such as negative bias temperature instability (NBTI)2,3 and data retention properties in non-volatile memory.4,5 However, a lack of methods for hydrogen concentration measurements using microscopy has veiled the degradation mechanism of device characteristics. Conventional methods such as secondary ion mass spectrometry (SIMS),6,7 nuclear reaction analysis (NRA)8 and elastic recoil detection analysis (ERDA)9 are successfully used to analyze hydrogen concentrations and distributions for blanket structures or structures larger than 100 μm. However, the three-dimensional (3D) spatial resolution of these methods is not sufficient for recent devices. Moreover, the accurate quantification of hydrogen has been very difficult because hydrogen atoms in the bulk or at the surface possibly diffuse during the analysis, or the amount of hydrogen atoms adsorbed from residual gas in the vacuum chamber onto the specimen surface is not always negligible.10 

Atom probe tomography (APT) is a practical method to obtain 3D elemental distributions in materials with almost atomic-scale resolution.11,12 Recent developments of ultraviolet laser-assisted APT have enabled the analysis of not only metals but also semiconductors13–16 and insulators. APT is also applicable to actual device structures, such as metal-oxide-semiconductor field-effect transistors (MOSFETs) in addition to blanket structures.17–21 

Hydrogen atoms in metallic materials, semiconductors and ceramics have been studied using APT.22–24 However, hydrogen atoms in insulating thin films have not been investigated. Moreover, previous reports were concerned with the imaging of hydrogen atoms segregated at specific sites such as grain boundaries or interfaces, and no quantification has been conducted. The dependence of hydrogen concentrations on experimental APT conditions has also been investigated with focus on the adsorption of hydrogen atoms from residual gas; however, the results have been insufficient to provide clear understanding for quantification.25 

In this study, quantification of hydrogen concentration in an insulating thin film, a SiN thin film in a SiO2/SiN/SiO2 (ONO) stack structure, is demonstrated using ultraviolet laser-assisted APT. The amount of hydrogen adsorbed from the residual gas is estimated by analyzing another ONO specimen with an extremely low hydrogen concentration of less than 0.2 at. % estimated by SIMS. In addition, the accuracy of quantification using APT is evaluated by comparison with the result obtained by ERDA.

Two types of ONO samples were prepared in this study. The bottom-SiO2 layer with a thickness of approximately 5 nm was formed by thermal oxidation of a Si substrate, and a SiN film with a thickness of approximately 6 nm was deposited by low-pressure chemical vapor deposition (LPCVD). The top-SiO2 layer with a thickness of approximately 8 nm was deposited by LPCVD on the SiN layer. A 60 nm thick poly-Si layer was then formed. In sample A, prepared as an ONO with an extremely low hydrogen concentration, spike annealing at a temperature of 1000 °C was conducted after deposition of the poly-Si layer to reduce the hydrogen concentration in the SiN layer. In sample B, which was prepared as an ONO with higher hydrogen concentration, a 20 nm thick cap-SiN layer was deposited by LPCVD onto the poly-Si layer before spike-annealing to suppress the out-diffusion of hydrogen atoms. Details of the sample fabrication procedure are described elsewhere.26 

Figure 1 shows SIMS depth profiles of hydrogen for samples A and B. In sample B, three peaks are observed in the depth profile. The peak at the top side is attributed to the poly-Si/top-SiO2 interface, the central peak is attributed to the SiN layer, and the peak at the bottom side is attributed to the bottom-SiO2/Si-substrate interface.26 In sample A, the peak height in the SiN layer was much lower than that in sample B, as was intended.

FIG. 1.

SIMS depth profiles of hydrogen for samples A and B.

FIG. 1.

SIMS depth profiles of hydrogen for samples A and B.

Close modal

Samples were milled into needle shapes using a focused ion beam (FIB) before APT analysis. The needles were formed in the direction parallel to the interfaces to achieve a larger analysis volume.27 APT analysis was performed using a LEAP 4000X HR atom probe (Ametek). The detection efficiency of this system is approximately 37%, which is limited by geometrically open area of the detector system. This instrument is equipped with an ultraviolet pulsed-laser (wavelength 355 nm) system to assist field evaporation. The pulsed-laser frequency and energy were set to 200 kHz and 70 pJ, respectively, and the temperature of specimens was kept at 50 K. The evaporation rate, which is defined as detected ions/laser pulse, was kept at 0.4% or 0.8% by controlling the bias voltage at the specimen. The chamber pressure was kept between 2.5 × 10−9 Pa and 3.0 × 10−9 Pa for all measurements. The accuracy of APT quantification was checked by ERDA and Rutherford backscattering spectrometry (RBS; Kobelco HRBS500). RBS was conducted using a He+ ion beam with an energy of 450 keV and a current of 31 nA to measure oxygen and nitrogen depth profiles. ERDA was performed using an N+ ion beam with an energy of 480 keV and a current of 2 nA to obtain a hydrogen depth profile. The concentration of hydrogen was calibrated by analyzing a diamond-like carbon film with a known hydrogen concentration.

Figures 2(a) and 2(b) show 3D atom maps for samples A and B, respectively. The evaporation rate was initially set to 0.4 % and changed to 0.8 % as indicated in Figs. 2(a) and 2(b) to investigate the effect of hydrogen adsorption. As we intended, the concentration of hydrogen atoms in the SiN layer for sample A was smaller than that for sample B. Figures 2(c) and 2(d) show depth profiles of 1H+, 1H2+, oxygen (including isotopes and molecules) and 28Si14N+ plotted perpendicularly to the interfaces in 0.5 nm steps. Ionic concentration is defined as the number of specific ions / total number of detected ions. Other molecular ions including hydrogen such as 28Si1H+ were not estimated because the mass resolution M/ΔM, which was approximately 500, was inadequate to distinguish with other peaks such as 29Si+. The average concentrations of 1H+ ions in the SiN layer for samples A and B were approximately 0.7 and 1.6 ionic %, while the average concentrations of 1H2+ were approximately 0.11 and 0.07 ionic %, respectively. The ratio of hydrogen concentrations between the two samples was smaller than that expected from the SIMS analyses. The concentration of 1H+ in the SiO2 layer ranged from 0.6 to 1.1 ionic %.

FIG. 2.

3D atom maps of ONO measured using APT for samples (a) A and (b) B, and depth profiles of 1H+, 1H2+, oxygen and 28Si14N+ for samples (c) A and (d) B.

FIG. 2.

3D atom maps of ONO measured using APT for samples (a) A and (b) B, and depth profiles of 1H+, 1H2+, oxygen and 28Si14N+ for samples (c) A and (d) B.

Close modal

Seven regions of interest (ROIs) were set as shown in Figs. 3(a) and 3(b). The evaporation rate was changed from 0.4% to 0.8% at the border between ROI 3 and ROI 4. Figures 3(c) and 3(d) show hydrogen concentrations in the SiN, top-SiO2, and bottom-SiO2 layers, respectively, at each ROI for samples A and B. It is clear that the hydrogen concentrations (both 1H+ and 1H2+) in both SiN and SiO2 layers gradually increased except from ROI 3 to ROI 4. As the analysis proceeded, the tip radius became gradually larger, and the evaporation rate per unit area decreased. Consequently, the increase of hydrogen during the analysis indicates that hydrogen atoms adsorbed onto the sample surface from the residual gas gradually increased. The similar phenomenon was reported in Ref. 25.

FIG. 3.

ROIs in samples (a) A and (b) B, and concentration of hydrogen atoms in the SiN, top-SiO2 and bottom-SiO2 layers at each ROI for samples (c) A and (d) B.

FIG. 3.

ROIs in samples (a) A and (b) B, and concentration of hydrogen atoms in the SiN, top-SiO2 and bottom-SiO2 layers at each ROI for samples (c) A and (d) B.

Close modal

The hydrogen concentration became smaller when the evaporation rate was switched from 0.4% to 0.8%, i.e. ROI 3 to ROI 4, as shown in Figs. 3(c) and 3(d). When the evaporation rate was set higher, the electric field near the surface became stronger and the evaporation rate per unit area increased. Consequently, this decrease of hydrogen concentration indicates that adsorbed hydrogen decreased. Therefore, our results suggest that the adsorption is a dominant factor for the change of hydrogen concentration during the analysis.

To estimate the hydrogen concentration in the SiN layer, the concentration of hydrogen adsorbed from the residual gas should be distinguished from the concentration of hydrogen present in the SiN film. We focused on sample A, in which the concentration of hydrogen atoms present in the SiN film was negligible, because most of the detected hydrogen atoms could be regarded as being adsorbed from residual gas. Figure 4 shows the relationship between the hydrogen concentrations detected in the SiO2 and SiN layers at each ROI for several measurements. One plot is corresponding to the data from one ROI. The rectangle plots are results from three APT specimens of sample A, and the circle plots are results from three APT specimens of sample B. The relationship in Fig. 4 was almost linear and was maintained at every ROI and every measurement for both samples A and B. This result indicates that reproducible APT results can be obtained by investigating this relationship. The plots were fitted using the linear least squares method and the results were shown as the dashed lines in Fig. 4. We estimated the actual hydrogen concentration present in the SiN layer for sample B as 1.0 ± 0.2 (3σ) ionic % by taking a difference between the fitted lines. The atomic concentration of hydrogen is also approximately 1.0 ± 0.2 (3σ) at. % because the effect of molecular ions is trivial.

FIG. 4.

Relationship between the hydrogen concentrations detected in the SiO2 (mean concentrations in top-SiO2 and in bottom-SiO2) and SiN layers at each ROI. The rectangle plots are results from three APT specimens for sample A, and the circle plots are results from three APT specimens for sample B. Plots with the same color indicate results from the same specimen but from different ROIs. There are 7, 4 and 1 ROIs in sample A, and 7, 2 and 2 ROIs in sample B.

FIG. 4.

Relationship between the hydrogen concentrations detected in the SiO2 (mean concentrations in top-SiO2 and in bottom-SiO2) and SiN layers at each ROI. The rectangle plots are results from three APT specimens for sample A, and the circle plots are results from three APT specimens for sample B. Plots with the same color indicate results from the same specimen but from different ROIs. There are 7, 4 and 1 ROIs in sample A, and 7, 2 and 2 ROIs in sample B.

Close modal

This estimation is derived from only 1H+ ions, and molecular ions such as 1H2+ and 28Si1H+ were ignored. Most of 1H2+ is originated from adsorption because the concentrations of 1H2+ between samples A and B were almost the same as shown in Figs. 3(c) and 3(d). The contribution of 28Si1H+ is impossible to be fixed because of the lack of the mass resolution to distinguish 29Si+ and 28Si1H+. Moreover, it is not available from the natural isotope ratio of 28Si+ and 29Si+ because 28Si+ and 14N2+ is not distinguished. However, the ionic concentration of 28Si1H+ is roughly estimated as approximately 0.1% from the ratio of 30Si+ and 30Si1H+ whose ratio is not largely affected by the other silicon isotopes and silicon hydrides. Thus, the effect of ignoring the contribution of the molecular ions is relatively small and the accuracy of the quantification is maintained.

Figure 5 shows depth profiles of nitrogen, oxygen and hydrogen analyzed using RBS and ERDA for sample B, of which the hydrogen concentration in the SiN layer determined using APT was 1.0 at. %. The hydrogen concentration in the SiN layer measured from ERDA was approximately 1.7 ± 1.0 at. %, which is consistent with the result obtained from APT. Thus, the accuracy of APT quantification was confirmed.

FIG. 5.

Depth profiles of nitrogen, oxygen and hydrogen analyzed using RBS and ERDA for sample B.

FIG. 5.

Depth profiles of nitrogen, oxygen and hydrogen analyzed using RBS and ERDA for sample B.

Close modal

In summary, we have quantitatively estimated the concentration of hydrogen atoms in a SiN thin film with an ONO structure using APT. These experiments indicate that the change in hydrogen concentration during the APT analysis could be explained by the adsorption of hydrogen atoms onto the specimen surface from residual gas in the vacuum chamber. The concentration of adsorbed hydrogen atoms was derived by analysis of an ONO with an extremely low hydrogen concentration. The accuracy of the APT quantification was confirmed by comparison with the ERDA quantification. The present results indicate that APT can be used to image 3D distributions of hydrogen atoms in actual device structures at almost the atomic scale, which have not been achieved with other methods.

This work at Tohoku University was supported in part by JSPS KAKENHI Grant Numbers 26289097 and 15H05413.

1.
International Technology Roadmap for Semiconductors 2013 edition, Front End Processes, http://www.itrs.net/.
2.
K. O.
Jeppson
and
C. M.
Svensson
,
J. Appl. Phys.
48
,
2004
(
1977
).
3.
S.
Mahapatra
,
D.
Saha
,
D.
Varghese
, and
P. B.
Kumar
,
IEEE Trans. Electron Dev.
53
,
1583
(
2006
).
4.
Z.
Liu
,
T.
Saito
,
T
Matsuda
,
K.
Ando
,
S.
Ito
,
M.
Wilde
, and
K.
Fukutani
, in
IEEE Proceedings of International Reliability Physics Symposium (IRPS2008)
, 705.
5.
S.
Fujii
and
K.
Sakuma
, in
IEEE Proceedings of International Reliability Physics Symposium (IRPS2013)
, 3B.5.1.
6.
R.
Gale
,
F. J.
Feigl
,
C. W.
Magee
, and
D. R.
Young
,
J. Appl. Phys.
54
,
6938
(
1983
).
7.
Y.
Kawashima
,
Z.
Liu
,
K.
Terashima
,
K.
Hamada
,
K.
Fukutani
,
M.
Wilde
,
S.
Aoyagi
, and
M.
Kudo
,
Appl. Surf. Sci.
212–213
,
804
(
2003
).
8.
J.
Krauser
,
F.
Wulf
,
M. A.
Briere
,
J.
Steiger
, and
D.
Braeunig
,
Microelectron. Eng.
22
,
65
(
1993
).
9.
M.
Ionescu
,
B.
Richards
,
K.
McIntosh
,
R.
Siegele
,
E.
Stelcer
,
O.
Hawas
,
D.
Cohen
, and
T.
Chandra
,
Mater. Sci. Forum
539-543
,
3551
(
2007
).
10.
D. E.
Barofsky
and
E.W.
Mueller
,
Surf. Sci.
10
,
177
(
1968
).
11.
T. F.
Kelly
and
M. K.
Miller
,
Rev. Sci. Instrum.
78
,
031101
(
2007
).
12.
G. L.
Kellogg
and
T. T.
Tsong
,
J. Appl. Phys.
51
,
1184
(
1980
).
13.
Y.
Shimizu
,
Y.
Kawamura
,
M.
Uematsu
,
M.
Tomita
,
T.
Kinno
,
N.
Okada
,
M.
Kato
,
H.
Uchida
,
M.
Takahashi
,
H.
Ito
,
H.
Ishikawa
,
Y.
Ohji
,
H.
Takamizawa
,
Y.
Nagai
, and
K. M.
Itoh
,
J. Appl. Phys.
109
,
036102
(
2011
).
14.
K.
Hono
,
T.
Ohkubo
,
Y. M.
Chen
,
M.
Kodzuka
,
K.
Oh-ishi
,
H.
Sepehri-Amin
,
F.
Li
,
T.
Kinno
,
S.
Tomiya
, and
Y.
Kanitani
,
Ultramicroscopy
111
,
576
(
2011
).
15.
N.
Dawahre
,
G.
Shen
,
S.
Balci
,
W.
Baughman
,
D.
Wilbert
,
N.
Harris
,
L.
Butler
,
R.
Martens
,
S.
Kim
, and
P.
Kung
,
J. Electron. Mater.
41
,
801
(
2012
).
16.
D. J.
Larson
,
R. L.
Alvis
,
D. F.
Lawrence
,
T. J.
Prosa
,
R. M.
Ulfig
,
D. A.
Reinhard
,
P. H.
Clifton
,
S. S. A.
Gerstl
,
J. H.
Bunton
,
D. R.
Lenz
,
T. F.
Kelly
, and
K.
Stiller
,
Microsc. Microanal.
14
,
1254
(
2008
).
17.
K.
Inoue
,
F.
Yano
,
A.
Nishida
,
H.
Takamizawa
,
T.
Tsunomura
,
Y.
Nagai
, and
M.
Hasegawa
,
Ultramicroscopy
109
,
1479
(
2009
).
18.
K.
Inoue
,
F.
Yano
,
A.
Nishida
,
T.
Tsunomura
,
T.
Toyama
,
Y.
Nagai
, and
M.
Hasegawa
,
Appl. Phys. Lett.
92
,
103506
(
2008
).
19.
H.
Takamizawa
,
K.
Inoue
,
Y.
Shimizu
,
T.
Toyama
,
F.
Yano
,
T.
Tsunomura
,
A.
Nishida
,
T.
Mogami
, and
Y.
Nagai
,
Appl. Phys. Express
4
,
036601
(
2011
).
20.
K.
Inoue
,
F.
Yano
,
A.
Nishida
,
T.
Tsunomura
,
T.
Toyama
,
Y.
Nagai
, and
M.
Hasegawa
,
Appl. Phys. Lett.
93
,
133507
(
2008
).
21.
B.
Han
,
H.
Takamizawa
,
Y.
Shimizu
,
K.
Inoue
,
Y.
Nagai
,
F.
Yano
,
Y.
Kunimune
,
M.
Inoue
, and
A.
Nishida
,
Appl. Phys. Lett.
107
,
023506
(
2015
).
22.
D.
Haley
,
S. V.
Merzlikin
,
P.
Choi
, and
D.
Raabe
,
Int. J. Hydrogen Energy
39
,
12221
(
2014
).
23.
G.
Sundell
,
M.
Thuvander
,
A. K.
Yatim
,
H.
Nordin
, and
H.-O.
Andren
,
Corros. Sci.
90
,
1
(
2015
).
24.
H.
Takamizawa
,
K.
Hoshi
,
Y.
Shimizu
,
F.
Yano
,
K.
Inoue
,
S.
Nagata
,
T.
Shikama
, and
Y.
Nagai
,
Appl. Phys. Express
6
,
066602
(
2013
).
25.
G.
Sundell
,
M.
Thuvander
, and
H.-O.
Andren
,
Ultramicroscopy
132
,
285
(
2013
).
26.
Y.
Sakurai
,
Y.
Kunimune
,
M.
Inoue
,
Y.
Maruyama
,
A.
Nishida
, and
T.
Ide
,
Jpn. J. Appl. Phys.
53
,
08LC03
(
2014
).
27.
S.
Koelling
,
N.
Innocenti
,
G.
Hellings
,
M.
Gilbert
,
A. K.
Kambham
,
K.
De Meyer
, and
W.
Vandervorst
,
Ultramicroscopy
111
,
540
(
2011
).