Measurements of RF noise in InGaAs/InAlAs recessed diodes:

Signatures of shot-noise suppression

1

O. García-Pérez, J. Mateos, S. Pérez, T. González A. Westlund, J. Grahn

2

Shot Noise

Uncorrelated carriers

Correlated carriers Shot-noise SI=qI

full shot noise = 1

suppressed shot noise 1enhanced shot noise 1

Origin of correlations: Pauli exclusion principle (degenerate semiconductors) Long-range Coulomb interaction (strong space-charge effects)

Fano Factor = SI qI

The measurement of shot noise and the value of its corresponding Fano Factor can provide valuable insight about the transport dynamics

inside semiconductor devices

Shot-noise SI=qI is related to the discrete character of the electronic charge and is usually observed in electronic devices when carrier transport is ballistic or

is limited by an energy barrier (Schottky diodes, tunnel diodes…)

Javier Mateos – Universidad de Salamanca

Shot Noise

x / L0.0 0.2 0.4 0.6 0.8 1.0

Pote

ntia

l (K

BT/

q)

-5

0

5

10

15 qU/KBT

4020

10

4

0

80

nc nc

thermalequilibrium

thermalequilibrium

x=L

N

x=0

Active region

Contact 1 Contact 2

UPoissonian

injectionPoissonian

injection

Ballistic diodeShot noise suppression due to

long range Coulomb correlations

high important space-charge effects

DcLL

Full shot noise is recovered when the barrier disappears

Presence of a barrier that disappears when voltage is increased

3Javier Mateos – Universidad de Salamanca

4

x / L0.0 0.2 0.4 0.6 0.8 1.0

Pote

ntia

l (K

BT/

q)

-5

0

5

10

15 qU/KBT

4020

10

4

0

80

nc nc

thermalequilibrium

thermalequilibrium

x=L

N

x=0

Active region

Contact 1 Contact 2

UPoissonian

injectionPoissonian

injection

Position(um)

0.0 0.1 0.2 0.3 0.4 0.5 0.6P

oten

tial (

V)

-0.4-0.20.00.20.40.60.81.01.21.41.61.8

Ballistic diode

Presence of a barrier that disappears when voltage is increased

Shot NoiseRecessed planar diodes=Slot diodes

(ungated HEMTs)

Shot noise suppression expected

Quasi-ballistic transport in the channel

Potential barrier imposed by the surface

charges at the recess

Javier Mateos – Universidad de Salamanca

Source DrainRecess

LRLS LD

capInGaAs

InGaAsInAlAs

InAlAsChannel

5

Slot diodes: fabrication

Parameter Reference design [nm]

Tested range [nm]

Ls 200 200-800

Lr 200 200-800

Ld 550 300-1000

Slot diodes with different geometries have been fabricated and characterized

2-8 nm

Diodes without recess+

Javier Mateos – Universidad de Salamanca

6

Slot diodes: Monte Carlo simulations

Cap layer

Source DrainLs Lr Ld

Applied voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Cur

rent

den

sity

(x10

3 A/m

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

LR=160 nm (Exp.)LR=400 nm (Exp.)LR=800 nm (Exp.)LR=160 nm (MC)LR=400 nm (MC)LR=800 nm (MC)

Applied voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Cur

rent

den

sity

(x10

3 A/m

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

LR=160 nm (Exp.)LR=400 nm (Exp.)LR=800 nm (Exp.)LR=160 nm (MC)LR=400 nm (MC)LR=800 nm (MC)

Batch C7, type 6 - Cap 2nm

Applied voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Cur

rent

den

sity

(x10

3 A/m

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

LR=160 nm (Exp.)LR=400 nm (Exp.)LR=800 nm (Exp.)LR=160 nm (MC)LR=400 nm (MC)LR=800 nm (MC)

Batch C7, type 2 - Cap 4nm Batch C7, type 5 - Cap 8nm

Ls = 200nmLd = 550nm160nm<Lr<800nm

Contact Resistance

(0.35 Ω.mm)+

MC simulations reproduce the experimental DC curves and their dependence on Ls, Lr and Ldby adjusting the values of the surface charges and including the ohmic contact resistances

Javier Mateos – Universidad de Salamanca

7

MC simulations reproduce the experimental DC curves and their dependence on Ls, Lr and Ldby adjusting the values of the surface charges and including the ohmic contact resistances

Cap layer

Source DrainLs Lr Ld Contact Resistance

(0.35 Ω.mm)+

Ls = 200nmLr = 160nm300nm<Ld<1000nm

Applied voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Cur

rent

den

sity

(x10

3 A/m

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

LD=300 nm (Exp.)LD=550 nm (Exp.)LD=1000 nm (Exp.)LD=300 nm (MC)LD=550 nm (MC)LD=1000 nm (MC)

Applied voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Cur

rent

den

sity

(x10

3 A/m

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

LD=300 nm (Exp.)LD=550 nm (Exp.)LD=1000 nm (Exp.)LD=300 nm (MC)LD=550 nm (MC)LD=1000 nm (MC)

Applied voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2

Cur

rent

den

sity

(x10

3 A/m

)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

LD=300 nm (Exp.)LD=550 nm (Exp.)LD=1000 nm (Exp.)LD=300 nm (MC)LD=550 nm (MC)LD=1000 nm (MC)

Batch C7, type 6 - Cap 2nm Batch C7, type 2 - Cap 4nm Batch C7, type 5 - Cap 8nm

Slot diodes: Monte Carlo simulations

Javier Mateos – Universidad de Salamanca

8

Noise measurements

Reference plane

Device

PNA-X N5244

VDC

S11 + Power density

Measurements performed on wafer with a VNA Agilent PNA-X N5244 (with dedicated receivers for high sensitivity noise power measurements) in the range between 20 and 30 GHz.

Javier Mateos – Universidad de Salamanca

9

Noise measurements

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Cur

rent

(mA

)

0

5

10

15

20LD=300nmLD=550nmLD=1000nmNo recess (1300nm)

Current (mA)0 2 4 6 8 10 12 14 16

SI(x

10-2

1 A2 /H

z)0.0

0.5

1.0

1.5

2.0

2.5

3.0

LD=300nmLD=550nmLD=1000nmNo recess (1300nm)2qI

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LD=300nmLD=550nmLD=1000nmNo recess (1300nm)

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Cur

rent

(mA

)

0

5

10

15

20LR=160nmLR=400nmLR=800nmNo recess (1300nm)

Current (mA)0 2 4 6 8 10 12 14 16

SI (

x10-2

1 A2 /H

z)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

LR=160nmLR=400nmLR=800nmNo recess (1300nm)2qI

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LR=160nm

LR=400nmLR=800nmNo recess (1300nm)

Ls = 200 nm, Lr = 160 nm, Ld = 300, 500, 1000 nm

Ls = 200 nm, Ld = 550 nm, Lr = 160, 400, 800 nm

x Non-recessed diode with L = 1.3 m

Full Shot Noise

Full Shot Noise

Javier Mateos – Universidad de Salamanca

10

Noise measurements

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Cur

rent

(mA

)

0

5

10

15

20LD=300nmLD=550nmLD=1000nmNo recess (1300nm)

Current (mA)0 2 4 6 8 10 12 14 16

SI(x

10-2

1 A2 /H

z)0.0

0.5

1.0

1.5

2.0

2.5

3.0

LD=300nmLD=550nmLD=1000nmNo recess (1300nm)2qI

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LD=300nmLD=550nmLD=1000nmNo recess (1300nm)

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Cur

rent

(mA

)

0

5

10

15

20LR=160nmLR=400nmLR=800nmNo recess (1300nm)

Current (mA)0 2 4 6 8 10 12 14 16

SI (

x10-2

1 A2 /H

z)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

LR=160nmLR=400nmLR=800nmNo recess (1300nm)2qI

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LR=160nm

LR=400nmLR=800nmNo recess (1300nm)

Ls = 200 nm, Lr = 160 nm, Ld = 300, 500, 1000 nm

Ls = 200 nm, Ld = 550 nm, Lr = 160, 400, 800 nm

x Non-recessed diode with L = 1.3 m

Noise level much lower than full shot noise level = shot noise suppression

Increase of noise at high bias, when current saturation starts

(not visible without recess)

Full Shot Noise

Full Shot Noise

Javier Mateos – Universidad de Salamanca

11

Noise measurements

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Cur

rent

(mA

)

0

5

10

15

20LD=300nmLD=550nmLD=1000nmNo recess (1300nm)

Current (mA)0 2 4 6 8 10 12 14 16

SI(x

10-2

1 A2 /H

z)0.0

0.5

1.0

1.5

2.0

2.5

3.0

LD=300nmLD=550nmLD=1000nmNo recess (1300nm)2qI

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LD=300nmLD=550nmLD=1000nmNo recess (1300nm)

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Cur

rent

(mA

)

0

5

10

15

20LR=160nmLR=400nmLR=800nmNo recess (1300nm)

Current (mA)0 2 4 6 8 10 12 14 16

SI (

x10-2

1 A2 /H

z)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

LR=160nmLR=400nmLR=800nmNo recess (1300nm)2qI

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LR=160nm

LR=400nmLR=800nmNo recess (1300nm)

Ls = 200 nm, Lr = 160 nm, Ld = 300, 500, 1000 nm

Ls = 200 nm, Ld = 550 nm, Lr = 160, 400, 800 nm

x Non-recessed diode with L = 1.3 m

Stronger shot noise suppression for longer Ld(and not depending on Lr)

Full Shot Noise

Full Shot Noise

Javier Mateos – Universidad de Salamanca

Noise models: discussion

Rc Ls Lr Ld

Ohmic region Thermal noise

Always at equilibrium Ss=4KBTRc+Rs

Ohmic region Thermal noise

Bias dependent resistance Sd=4KBTRd(V)

Barrier limited transport Shot noise Sr=F·2qI

Rc+Rs

Sr Ss Sd

Rr Rd(V)

12Javier Mateos – Universidad de Salamanca

Position(um)0.0 0.1 0.2 0.3 0.4 0.5 0.6

Pot

entia

l (V

)

-0.4-0.20.00.20.40.60.81.01.21.41.61.8

Noise models: discussion

Rc+Rs = 30

Ss=4KBTRc+Rs with T = 300 K

Rr = ? Si=F·2qI

Rd = ?

Sd=4KBT

Rdwith T = ?

Rc+Rs= 30 R

V

V

Si

Si=4kT0/Rc+Rs

Rc+Rs

Sr Ss Sd

Rr Rd(V)

Total resistance is measured RT(V)=Rc+Rs+Rr+Rd is known

The values of Rr and Rc can be estimated from

Monte Carlo simulations

Rr decreases and Rcstrongly increases when current saturates (due

to intervalley scattering)

Complex non-linear behavior Noise?

13Javier Mateos – Universidad de Salamanca

Noise models: discussion

Rc+Rs

Sr Ss Sd

Rr Rd(V)

Rr

R

V

V

Si

Si=g2qI

Rd

R

V

V

SiSi=4kT0/Rd(0)

Noise in Rd is initially supposed

to be constant (equilibrium value of Rd)

Rc+Rs = 30

Ss=4KBTRc+Rs with T = 300 K

Rr = ? Si=F·2qI

Rd = ?

Sd=4KBT

Rdwith T = ?

Rc+Rs= 30 R

V

V

Si

Si=4kT0/Rc+Rs

Simplified noise model

14

Full/suppressed shot noise in the active region

Simplified noise model

Javier Mateos – Universidad de Salamanca

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LR=160nmLR=400nmLR=800nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI (

10-2

1 A

2 /Hz)

0.0

0.5

1.0

1.5

2.0

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LD=300nmLD=550nmLD=1000nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI(1

0-21

A2 /H

z)

0.0

0.5

1.0

1.5

2.0

15

Noise measurements and models

LD=300nmLD=550nmLD=1000nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI(1

0-21

A2 /H

z)

0.0

0.5

1.0

1.5

2.0

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

LR=160nmLR=400nmLR=800nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI (

10-2

1 A

2 /Hz)

0.0

0.5

1.0

1.5

2.0

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

10

Ls = 200 nm, Lr = 160 nm, Ld = 300, 500, 1000 nm

Ls = 200 nm, Ld = 550 nm, Lr = 160, 400, 800 nm

Full Shot Noise

Full Shot Noise

Full shot noise in the active region

F=1

The simplified noise model provide good qualitative

agreement with the experimental results (that

could be improved by increasing the noise

temperature of both contact and drain resistances)

Javier Mateos – Universidad de Salamanca

Voltage (V)0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4

Fano

Fac

tor

0,1

1

Voltage (V)0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4

Fano

Fac

tor

0,1

1

Fano

Fac

tor

0,1

1

Voltage (V)0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Fano

Fac

tor

0.1

1

Voltage (V)0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4

Fano

Fac

tor

0,1

1

Fano

Fac

tor

0,1

1

LD=300nmLD=550nmLD=1000nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI(1

0-21

A2 /H

z)

0.0

0.5

1.0

1.5

2.0

LR=160nmLR=400nmLR=800nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI (

10-2

1 A

2 /Hz)

0.0

0.5

1.0

1.5

2.0

LR=160nmLR=400nmLR=800nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI (

10-2

1 A

2 /Hz)

0.0

0.5

1.0

1.5

2.0

LD=300nmLD=550nmLD=1000nm

Model Exp.

Current (mA)0 2 4 6 8 10 12

SI(1

0-21

A2 /H

z)

0.0

0.5

1.0

1.5

2.0

16

Noise measurements and modelsLs = 200 nm, Lr = 160 nm, Ld = 300, 500, 1000 nm

Ls = 200 nm, Ld = 550 nm, Lr = 160, 400, 800 nm

Full Shot Noise

Full Shot Noise

Completely suppressed shot noise in the active

region F=0

Full shot noise in the active region

F=1

Noise in the active region (and therefore the

signature of full/suppressed shot noise) is only visible on

the total noise at intermediate voltages

Noise at high bias depend just on the noise of the drain region

due to its high resistance

Javier Mateos – Universidad de Salamanca

• Measurements of noise performed in a set of recessed planar InGaAs/InAlAs diodes with different dimensions show potential signs of shot noise suppression in the structures due to the presence of a potential barrier

• A detailed analysis of the noise contribution of the different regions of the devices shows that contact, source and drain resistances strongly affect the value of the total noise

• The possible shot noise suppression appearing under the recess could just be visible on the total noise at intermediate bias (before the onset of intervalley scattering) where the resistance of the drain region is still low

• Devices with reduced access and drain resistances should be fabricated in order to obtain conclusive results

17

Conclusions

Javier Mateos – Universidad de Salamanca