Non-standard geometry scaling effects

Transcription

1 Non-standard geometry scaling effects S. Lehmann 1), M. Schröter 1),2), J. Krause 1), A. Pawlak 1) 1) Chair for Electron Devices and Integr. Circuits, Univ. of Technol. Dresden, Germany 2) ECE Dept., University of California San Diego, La Jolla, CA, USA HICUM Workshop, Heilbronn (Germany), 12./13. June 26 SL, MS 1

2 Outline Motivation for investigations profile Variations of the reference profile Results for electrical characteristics and FOMs (f T, I-V,...) Summary SL, MS 2

3 Motivation for investigation Designer s goal Modeller s goal Circuit optimization Device characterization Find optimal device Compact models for as many as possible devices and device configurations Measure a few devices Extract specific parameters Apply scaling equations! Measurements of some advanced process technologies show deviations w.r.t. conventional scaling equations. SL, MS 3

4 Motivation for investigation measurements of scaling of collector current symbols... measurements lines... linear regression measurements of scaling of breakdown voltage BV CE A E SL, MS 4

5 profile profile based on SIMS measurements (in the following => Standard): Non-selective epitaxy (Ge-profile lateral constant) Doping of the internal transistor is laterally constant Simulated emitter widths: b E =[ ]µm Field oxide E.5.1 B log( D ), N122C128 E D [cm -3 ] 1e+2 1e+19 1e+18 B C x [μm] e+17 1e+16 1e y [μm] x [µm] y [µm].3.35 => obeys standard scaling rules SL, MS 5

6 profile 1e e+2 1e+19 cutline at y=.4µm D [cm -3 ] 1e+18 1e e+16 1e+15 internal SIMS cutline at y=.µm Baselink SIMS internal DEVICE Baselink DEVICE Ge SIMS.5 Ge DEVICE x [µm].1 SL, MS 6

7 profile Gummel plot Transit frequency =V =V I C /A E b E b E =.25μm b E =.45μm 1 3 b =.65μm E b E =1.5μm b =1.45μm E V BE [V] f T [GHz] b E =.25μm b E =.45μm b E =.65μm b E =1.5μm b E =1.45μm I C /A E b E SL, MS 7

8 Depthvar1:.5 s E = Δb.2 Depthvar2:.5 s E = Δb Variations of the reference profile/1 Variations of doping with emitter width (b E,nom =.65µm): Base-emitter junction depth SIC width increases increases/decreases with emitter width with emitter width (Depthvar1/Depthvar2) (SICvar1) (b E,nom +Δb)/2 2/3x b E,nom /2 je 1/3x je Contour at D=cm 3 E x je,nom x je,nom *s E Sicvar1: with s E as Depthvar b SIC b SIC *s E (b E +Δb)/2 Doping along intersection line at depth x=.5μm b E /2 intersection at depth x=.5µm x [μm] y [μm] D [cm -3 ] D [cm 3 ] 1 17 b E.1.2 y [µm] SL, MS 8

9 conventional edge Variations of the reference profile/2 Increased edge doping due to fast poly edge diffusion (Polyvar) (b E <b E,nom ) (b E,nom ) (b E >b E,nom ) Non-standard lateral doping Transient enhanced diffusion (TEDvar) (b E <b E,nom ) (b E,nom ) TED edge (b E >b E,nom ) modified edge E log( D ), N122D conventional edge E log( D ), N122D x [μm] x [μm] y [μm] y [μm] 14 SL, MS 9

10 Results - Transit frequencies Large influence on transit frequencies for Depthvar1 and Depthvar2 For small emitter width Polyvar shows the same behavior as Depthvar2 (smaller neutral base width, higher doping at BE junction => larger BE junction capacitance) For small emitter width TEDvar shows the reverse effect as Polyvar (wider neutral base width, lower doping at BE junction => smaller BE junction capacitance) For larger emitter width the variations of the edge doping (Polyvar, TEDvar) show decreased deviations w.r.t. reference profile =V, b E =.25μm =V, b E =1.45μm f T [GHz] Depthvar1 Depthvar2 TEDvar Polyvar Sicvar1 Sicvar2 Sicvar3 f T [GHz] Depthvar1 Depthvar2 TEDvar Polyvar Sicvar1 Sicvar2 Sicvar3 x je w B C je x je w B C je I C /A E I C /A E SL, MS 1

11 Results - Collector current I C /A E 1.55 x Standard method for perimeter-area separation Polyvar and Sicvar1 still show linear scaling behavior Depthvar1 w B =V, V BE =.65V I C /A E 1.8 x Depthvar2 w B =V, V BE =.65V I C /A E P E /A E [1/μm] 1.55 x TEDvar =V, V BE =.65V w B 1.1 P /A [1/μm] E E I C /A E P /A [1/μm] E E 1.8 x 1 3 Polyvar slope ~I Cp =V, V BE =.65V 1.1 P E /A E [1/μm] SL, MS 11

12 I B /A E 4.2 x Results - Base current Standard method for perimeter-area component separation Sicvar1, Polyvar and TEDvar show linear behavior (with varying peripheral component) 3.75 P E /A E [1/μm] 1 x Depthvar1 TEDvar =V, V BE =.65V w E =V, V BE =.65V v ce =3*1 5 cm/s I B /A E 4.2 x Depthvar2 =V, V BE =.65V 3.75 P E /A E [1/μm] 4.8 x w E Polyvar =V, V BE =.65V I B /A E slope ~I Bp I B /A E slope ~I Bp P /A [1/μm] E E 3.6 P /A [1/μm] E E SL, MS 12

13 Results - Low-current forward charge Q f /A E [fc/μm 2 ] Depthvar1 Depthvar2 TEDvar Polyvar Sicvar1 determined at V BE =.75V.8 P E /A E [1/μm] Q f = τ f I C, Sicvar as well as Polyvar, TEDvar: the larger devices deviate from linear scaling slightly Depthvar1 (increasing junction depth with regard to emitter width) shows significantly more nonlinear scaling behavior than Depthvar2 SL, MS 13

14 Results - Junction capacitances C je /A E [ff/μm 2 ] Depthvar1 V BE =.2V x je N BE 7 x je N BE 6.5 C je /A E [ff/μm 2 ] Depthvar2 V BE =.2V P /A [1/μm] E E P /A [1/μm] E E TEDvar V BE =.2V Polyvar V BE =.2V C je /A E [ff/μm 2 ] x je w B N BE C je /A E [ff/μm 2 ] N BEp P E /A E [1/μm] 5.5 P /A [1/μm] E E SL, MS 14

15 Results - Junction capacitances C jc = C jci A SIC + C jcx A BCx = C jci A SIC + C jcx d BCx P BCx Use of the correct dimensions for and would shift P=>P A SIC A BCx.9.88 Sicvar1 =.2V P b SIC d BCx.86 P C jc /A SIC [ff/μm 2 ] A SIC A E P P BCx.74 P P BC,eff /A SIC SL, MS 15

16 Tetrode simulations Symmetric structure as shown below has been simulated Voltage difference of +1mV between base contact B1 and B2. V B2 V B1 R = with. ( = + )/2 2r + r b E l Bx SBi E = 1μm I B1 I B2 l E I B1 B1 E N122C128 B2 I B2 x x [μm] y [μm] C 2 SL, MS 16

17 Results - Tetrode simulations Relevant deviations with respect to linear scaling are found only for Depthvar => direct influence of the varying neutral base width For very small emitter width a deviation of Polyvar is expected due to the same reason =V Depthvar1 Depthvar2 TEDvar x je w B Polyvar Sicvar1 x je w B R [kω] b [μm] E SL, MS 17

18 Results - BV CE For Depthvar, Sicvar, Polyvar only slight deviations in BV CE are observed TEDvar shows a significant higher BV CE and a dependence on emitter width determined at V BE =.55V I B Depthvar1 Depthvar2 TEDvar Polyvar Sicvar1 B VCE [V] b E measured example V CB b E [μm] SL, MS 18

19 Summary Variation of base-emitter junction depth with emitter width shows a large deviation with respect to conventional scaling => in dependence of occurence of similar measurements a modification of existing scaling equations has to be considered depending on observed measurement trends Some deviations with respect to conventional scaling occur only at small emitter width or or influence but even the peripheral components (e.g. Polyvar, in parts TEDvar) Evaluation of measured curves can allow conclusions regarding lateral doping profile (e.g. BV CE => TEDvar) Reason for deviation from conventional scaling for the forward charge Q f for large devices has to be investigated Develop preprocessor to include non-standard scaling behavior in scalable model parameter generation 19