`INNOLUX CORP. v. PATENT OF SEMICONDUCTOR ENERGY
`LABORATORY CO., LTD.
`
`IPR2013-00064
`
`
`
`THE EFFECT OF CONTACT OVERLAP DISTANCE ON a-Si TFT
`PERFORMANCE
`
`1025
`
`SHUICHI UCHIKOGA, MASAHIKO AKIYAHA, TAKASHI KOIZUMI, MITSUSHI
`SUZUKI
`Research and Development Center, TOSHIBA Corp.,
`Saiwai-ku, Kawasaki, 210, Japan
`
`IKEDA AND KOUJI
`
`1, Komukai Toshiba-cho,
`
`ABSTRACT
`
`The gate/source overlap distance
`an
`is
`important factor
`in
`(AL,)
`fabricating self-aligned TFT with passivating layer. Six types of TFT were
`fabricated using thin intrinsic a-Si layers such as 20 nm,50 nm and 100 nm and
`a n+:a-Si or a n+:llc-Si were used as the contact layer. The least required
`gate/source overlap distance, ALc
`is
`the critical overlap where the TFT
`performance
`is
`not
`limited by
`the contact. This
`AL,, was determined
`experimentally. AL,, was found to be significantly affected by the intrinsic
`a-Si layer thickness and the AL,c can be small by depositing a thin intrinsic
`layer. The intrinsic layer thickness dependence of AL,, in the linear region
`can be explained
`from a existing model. However,
`the behavior
`in the
`saturation region suggests the need of another model, which will be discussed
`here. The variation of n* contact layer shown to have a small effect on ALW.
`In order to determine appropriate AL, for self-alignment TFT,
`the magnitude
`of field effect mobility, threshold voltage and drain current as well as AL,c
`must be considered.
`
`INTRODUCTION
`
`The self-alignment
`technique is
`important especially for
`liquid crystal display
`(LCD) manufacturing. This
`is because
`either in
`large area LCDs or in high definition LCDs, pattern
`misalignment
`fatally degrades picture quality. One of
`the
`effective methods to avoid deterioration of picture quality is
`the self-alignment of passivating SiNx which covers the channel
`region
`the gate electrode. Passivating SiNx can be self-
`to
`aligned to the gate electrode by exposing photoresist from the
`back side of the substrate. Channel
`length can be fabricated
`uniformly without any mask alignment by using this technique.
`Hence, gate/source overlap (AL,)
`is made uniformly over the whole
`display area. Usually the overlap will be less than one micron.
`Furthermore,
`this small gate/source overlap reduces parasitic
`capacitance. However, small AL, limits TFT performance [11], [2].
`In
`this work,
`the authors wish to study factors which
`determine the least required AL,, which will be defined as AL,,.
`The TFT structure used
`in
`this work
`is
`shown
`in Fig.l. Two
`parameters, namely,
`intrinsic layer thickness and n+ contact
`layer quality, were varied to investigate the effect of AL,C on
`TFT characteristics.
`Lsvd
`
`Lovs
`
`DRAIN
`
`.............
`_____ _
`
`S-lOURCE
`RCE
`Fig. l; Cross sectional view of TFT
`with passivating layer.
`
`PASSIVATING SiNx
`
`Mat. Res. Soc. Symp. Proc. Vol. 258. P1992 Materials Research Society
`
`
`
`1026
`
`EXPERIMENTAL
`
`Device fabrication
`
`The Mo-Ta gate electrode was patterned on a glass substrate,
`and PECVD was utilized to deposit SiOx and SiNx as
`the gate
`insulators, a-Si as intrinsic layer, and SiNx as the passivating
`layer. The photomask was designed to vary AL,
`from
`-3 pm to 3
`pm and AL, was fixed to 3pm. Here, minus means an offset TFT.
`Intrinsic a-Si layers were deposited with the thickness of 20
`nm, 50 nm and 100 nm. Furthermore, phosphorus doped n+ amorphous
`silicon (n+:a-Si) or n+ microcrystalline silicon (n+:Pc-Si) were
`deposited as the contact layer. As a result, 6 types of TFT were
`fabricated. The channel length and width of the TFT were 12 pm
`and 60 pm, respectively.
`In this work,
`the main concern was focused on the source
`contact. L,,. and L,,, were designed to be constant,
`since
`it
`greatly affects device performance. From the measurement, L.,, and
`in a length about 2.5 pm.
`L,,d were
`
`Characterization
`
`All AL, were measured by using pictures taken from FE-SEM
`observation. TFT performance was characterized by field effect
`mobility (tin),
`threshold voltage (Vth), and drain current (Ids).
`pn and Vth were obtained for both linear region and saturation
`region. Drain voltage Vds=O.l[VI was used for the linear
`region
`and Vds=15[V]
`for
`the saturation
`region. Gradual
`channel
`approximation was adopted to obtain pn and Vth for linear region
`and saturation
`region.
`Ids at Vgs=15[V] was
`employed
`as
`Ids for both the linear
`characteristic
`and saturation region.
`The n+ contact layers were characterized by measuring
`its
`dark conductivities.
`The conductivities were measured
`by
`depositing approximately 1 pm thick phosphorus doped a-Si film.
`The conductivities were 1.0xl0 3
`(ohm-cm)- and 8.OxlOI (ohm-cm)1
`for a-Si and pc-Si, respectively.
`
`RESULTS
`
`Figures
`4
`and
`3
`2,
`the AL, dependence of TFT
`show
`performance for three different intrinsic
`layer thickness
`in
`terms of pn, Vth and drain current Ids(Vgs=15[V]). These results
`are obtained by TFTs with n':pc-Si contact layer. The figures
`clearly show that pn and Ids decrease as AL, reduces. The figures
`also show that pn and Ids become independent of AL, in the region
`L, >2 pm.
`The pn in AL,>2 pm greatly depends on the intrinsic layer
`thickness in
`the saturation
`region while the thickness does not
`affect the pn in
`the linear region, as it can be seen in Figs.2
`and 3. That is,
`in
`the saturation region (Fig.3),
`the average pn
`and 0.78 (cm 2 /V.S)
`at AL,>2 tm are 0.52 (cmZ/V.S), 0.63 (cmZ/V-S),
`for
`intrinsic
`layer
`thickness
`20
`nm,
`50 nm, and
`100 nm,
`(cm 2 /V-S),
`respectively.
`In
`the
`linear
`region,
`0.73
`0.73
`(cm/V-S), and 0.64 (cm 2/V.S)
`for thickness 20 tim, 50 nm,
`and 100
`nm, respectively. The drain current in
`the saturation
`region is
`also affected by the film thickness (Fig.4).
`
`
`
`Q8
`OL7
`- Q6
`Un
`>05
`0~4
`
`02
`
`01
`
`16
`14
`
`512
`
`I
`
`I
`
`o 20nm
`A 50cM
`0 00nm
`
`-
`
`1
`
`3
`
`I
`
`00/
`
`1
`
`-2
`
`I
`
`-I
`
`0
`ALs [pm]
`
`I
`
`I
`
`I
`
`2
`
`I
`I
`I-layer thldk
`0 20nm
`A 50cm
`0100 nm
`LmsMI Rim
`
`SN
`it
`
`-- o--
`
`A
`
`A
`
`0 A0 0
`A
`
`0
`
`0
`
`0-
`
`0
`
`1027
`
`7
`
`-0
`
`07 Ityrtkness
`0 20
`-m
`A 5omnm
`0)C3100 nm
`L5-SAmmai REmoN m.~go0
`
`02
`
`0.1
`
`
`
`IO I
`
`I
`
`0
`
`0
`
`-3 -2
`
`-1
`tsL5
`
`0
`(pum]
`
`1
`
`2
`
`3
`
`o 2Onm
`A 5Onto
`0
`0 1
`0nm
`SATL ToH R s io
`cn :pc-SI
`
`A
`
`a A0 0
`
`a a
`
`<a
`
`0
`
`0
`
`0
`
`14
`
`12
`10
`
`6-
`
`4
`
`2
`
`0000
`aM0-CO--_.bE
`
`-3
`
`-2
`
`I
`
`2
`
`3
`
`0
`-1
`ALs [pm]
`Fig.3; ALs dependence of
`field
`and
`effective mobility
`(above)
`the
`threshold voltage (below)
`in
`saturation
`region
`for n+: .c-Si
`contact layer.
`
`0.8
`0.7-
`
`'0
`
`A AzA
`
`-
`
`E
`
`0.3
`
`3
`
`A
`
`A
`
`I-ALsc (pn)
`
`0.2-
`0 .1- ,
`'3 -2
`
`-I
`
`1
`
`2
`
`3
`
`0
`(cid:127)C5Em]
`Fig.5; Definition of ALsc(pn),
`the least required ALs
`which
`is
`for field effective mobility.
`
`Fig.4; ALs dependence of the drain
`current at Vgs=1.5[Vl, Vds=0.l[V]
`(above) and Vds=15Vj (below) for
`n+:pc-Si contact layer.
`
`2
`
`-3
`
`I
`, I
`-2
`
`-0-
`0- OU
`
`I
`-
`-1
`
`--
`0
`IALs (pm]
`
`w--I--
`i
`-
`
`2
`
`3
`
`Fig.2; ALs dependence of field
`effective mobility
`(above)
`and
`threshold voltage (below)
`in
`the
`linear region for n+:pc-Si contact
`layer.
`
`4 3 2 1
`
`-3
`
`-2
`
`E3
`3-
`.:P
`
`i.
`
`lIntrins loyer
`thiclmma
`o 20nm
`A S0nm
`o IO0nm
`W/L 60/12
`2- Vd.15 VI
`
`IntrinsIc loyer
`o 2Onm
`A 50nm
`
`W/L- 60/12
`Vd.-O.I IV]
`VC. 15 IV]
`
`S(cid:127)
`
`-1
`
`0
`ALs [pm)
`
`I
`
`2
`
`3
`
`0
`
`6
`
`00
`
`0
`
`-3
`
`-1
`
`-2
`
`0
`ALs (pm]
`
`1
`
`2
`
`3
`
`
`
`1028
`
`Definition of ALse
`
`the TFT
`the effect of AL, on
`to evaluate
`In orlder
`the critical length, AL.I, will be introduced. The
`performance,
`required gate/source overlap
`meaning of AL,5 is
`the
`least
`distance, that does not deteriorate TFT performance.
`To show the procedure of obtaining AL,,, Fig.5 will be used
`in the case of pn. The values of pn and Ids at AL,>2 pm will be
`used as TFT characteristics which are not deteriorated by the
`gate/source contact overlap.
`The embodiment of obtaining ALI, experimentally is as shown
`below.
`I. Calculate the average pn (pn(avg)) and standard deviation(o).
`from the measured data point
`2. Define the minimum AL, as AL,,
`which satisfies pn(avg)-3o. Then, define the next smaller
`data point as AL,,.1 .
`3. Then we define AL, , for pn as
`
`In the same manner, AL,1 for'drain current (AL,,(lds))
`as
`
`is defined
`
`AL,, (Ids)= (AL,, ( Ids) -AL,,., (Ids) )/2.
`
`ALsc dependence of TFT performance
`
`Figures 6 and 7 are obtained by using the above definition
`layer thickness and for each contact layer.
`for each intrinsic
`is
`for the drain
`the ALI, dependence for pn and Fig.7
`is
`Fig.6
`layer
`that AL,, decreases as
`intrinsic
`It
`is clear
`current.
`that
`shown
`is
`reduces for both pn and
`Ids.
`It
`thickness
`the major factors which
`is one of
`layer thickness
`intrinsic
`determine AL,,.
`indicate that AL, can be
`from Figs.6 and 7
`These
`results
`alignment of
`layers when self
`made smaller for thinner intrinsic
`it must
`passivating SiNx by gate pattern is concerned. However,
`I
`I
`
`1.0-
`
`F
`
`40A
`
`A
`
`0
`E
`_1IIII<
`,_J
`
`0A
`
`EA
`
`A nc pc
`SATURAMT REGION
`* n*pC LINER REGION
`0
`0 n o-SI SATrIO REGo=
`a2
`S nto-SI LmNEo REGION
`
`~0
`
`S(cid:127)
`
`0
`
`0
`
`A e-pE
`Softwm~. Rama,
`A n*pc
`LaUm Resso
`o n
`-SI SA,.T.Anow RomA
`* n-a-SI LyAm, REGm,
`
`-1.0 k
`
`100
`50
`I-layer thickness (nm]
`Fig.6; Intrinsic layer thickness
`dependence of ALsc
`for
`field
`effective mobility.
`
`0b
`
`-1
`
`0
`
`0
`
`I-layer thickness (nm]
`Fig.7; Intrinsic layer thickness
`for drain
`of ALsc
`dependence
`current.
`
`
`
`1029
`
`be noted that AL,€(pn)
`and AL,,(Ids) do not correspond
`to each
`other.
`It
`is essential
`to decide which AL,, is appropriate
`for
`TFT ffbrication. Furthermore,
`the absolute value of pn and Ids
`must be cbnsidered, because the definition of AL,, is not related
`to the absolute values. These facts show that AL,, cannot be
`determined by a single factor. The required specifications for
`the TFT are also an important factor which determines the exact
`AL,5 .
`
`n' contact dependence on the TFT performance
`
`As shown
`in Figs.6 and 7,
`there appeared no significant
`dependence on the kind of contact layer used in
`the experiment.
`Referring to Kanicki
`[31,
`two kinds of n' contact layer used in
`the experiment with Mo deposited should have
`three orders
`difference in the contact resistance. The average Ids at ALs >
`2 pm for n' :a-Si did not differ much from Fig.4. The actual
`values were 4.OxlO"[A]
`(20 nm), 4.8xI0I[A)
`(50 nm) and 5.OxlO0
`'[A]
`for linear region, and 2.7x10"[A]
`(20 nm), 3.0xl0"'[A1
`(50
`nm)and 3.7xlO"1
`[A] for saturation region.
`Since the n' contact layer is as thin as 50 nm,
`there is a
`possibility of another
`factor which determines
`the contact
`resistance in the actual TFT.
`
`DISCUSSION
`
`Several models were proposed
`to obtain AL,,
`theoretically
`I1], 121). Models are based on
`(for e.g.
`two effects,
`namely,
`the
`field effect mobility and SCLC under the contacts. Here, we focus
`on the theory by Troutman et
`al.flJ.
`Even though AL,,(pn) is
`an
`important
`factor, pn is
`likely
`to differ
`in
`the way
`it
`is
`derived.
`In order to avoid arguments on the adequate method of
`obtaining lin, AL,,(Ids) will be discussed.
`According to Ref.[1J, TFT performance will be approximately
`the same as the semi-infinite contact when the overlap exceeds
`three times the characteristic
`length (denoted as A,). This will
`correspond to AL, ,:3A,
`this
`in
`paper. The solid line in Fig.7 is
`the calculated AL,, using the method
`in Ref.) 1]. It
`is apparent
`from Fig.7 that
`the calculated value shows agreement with the
`experimental data, especially in
`the linear region.
`This
`fit
`between
`the calculated AL,,(Ids)
`and
`the
`experimental data
`in
`the
`linear
`region can be explained as
`follows.
`In Fig.2, the pn is approximately the same for each film
`thickness in ALs>2uim where TFT performance is not limited by the
`contact. Therefore,
`it
`is easy to understand that the thinner'the
`intrinsic layer, a greater SCLC can be obtained. Hence,
`the ALsc
`decreases as the intrinsic layer thickness reduces.
`In
`the saturation region,
`it does not
`fit well as it
`is
`in
`the case of
`linear
`region. ALsc
`in
`the saturation
`region
`indicates a more gradual curve than the solid line. This suggests
`the need for adopting an additional effect to the existing model.
`One of the expected effects
`the fringing effect
`is
`caused by the
`overlap on top of the passivating SiNx
`(Le,
`in Fig.l)141,15].
`Fig.8 shows the effect
`of L,,. with the corresponding L,,d
`in
`the
`footnote of the figure. The effect of L0,,
`is significant in the
`saturation region while it hardly shows any effect in the linear
`region. Furthermore,
`the slope obtained for the saturation region
`differs according to the intrinsic layer thickness. Based on Shaw
`et al.[4],
`the carrier concentration at the interface between the
`
`
`
`1030
`
`largely affected by
`passivating SiNx and the intrinsic layer is
`reasonable to consider that
`Lee on the passivating SiNx.
`It
`is
`concentration varies depending on
`effect of L., on the carrier
`the intrinsic layer thickness and on the applied voltage to the
`the effect of L,, must be a strong
`drain electrode. Therefore,
`candidate for explaining intrinsic layer thickness dependence in
`the case of saturation region.
`
`0S
`
`(cid:127)0.6
`
`AL-3.9 [pm3
`
`-
`
`20
`50
`1 00
`
`1
`0
`A
`
`0
`0
`A
`
`1.77
`l1.59
`2.12
`
`0.2 -
`I 2 3 4 5 6 7
`L.ovs Eprn3
`
`CONCLUSION
`
`Fig.8; Lovs dependence of field
`___effective mobility.
`
`has been discussed
`AL,C
`required overlap
`least
`The
`intrinsic
`layer
`thickness
`AL,, decreases as
`experimentally.
`reduces, and significantly depend on intrinsic layer thickness.
`It has been shown for the linear region that the effect of the
`in
`is well explained by the theory
`intrinsic layer thickness
`Ref.[l], while the effect in the saturation region needs another
`the n+ contact
`model, possibly the fringing effect. Furthermore,
`layer dependence is not significant as far as this experiment is
`one of the most
`concerned. The
`intrinsic layer thickness
`is
`important factors in manufacturing self-alignment TFT.
`
`ACKNOWLEDGEMENT
`
`acknowledge Dr. Howard, Dr. Troutman
`The authors gratefully
`and Dr. Wisnieff of IBM Thomas J. Watson Research Center for
`useful discussions.
`
`REFERENCES
`
`[1] R.R.Troutman and A.Kotwal,IEEE Trans.Electron.Devices 36,
`2915
`(1989)
`[21 G.E.Possin, D.E.Castleberry, W.W.Piper, and H.G.Parks, Proc.
`of the SID 26/3, 183 (1985)
`[31 J.Kanicki, Appl.Phys.Lett., 53(20),1943(1988)
`[4] J.G.Shaw and M.Hack, J.Appl.Phys., 65(5),2124(1989)
`151 Y.Tanaka, K.Tsutsui, T.Koizumi, H.Yamamoto and T.Tsukata,
`Extended Abstracts of the 22nd Conf. on SSDM,1035(1990)
`
`