Characterization of Magnetic

Nanostructures for STT-RAM

Applications by use of Macro- and
Micro-scale Ferromagnetic Resonance

Tom Silva1, Hans Nembach1, Justin

Shaw1, Brian Doyle2, Kaan Oguz2,
Kevin O’brien2, and Mark Doczy2
1.National Institute of Standards and Technology, Boulder
2.Intel, Hillsboro
 NIST team members
Justin Shaw Hans Nembach
Nanomagnetics Project Nanomagnetics Project

Bob McMichael Martin Schoen Mike Schneider

NIST, Gaithersburg (Ph.D. Student, U. Regensburg) Spintronics Project

2
 The two faces of metrology

If you can’t measure it, you

can’t understand it…

Generalized
Specialized Understanding:
Understanding: Metrology
Reductio ad mathematicum Reductio ad profitus

3
 Overview

• STT-MRAM: Background and motivation

• VNA-FMR: Measuring damping and anisotropy in

blanket films.

• H-MOMM: Measuring damping and anisotropy in e-

beam patterned structures.

4
 STT-MRAM: a promising emerging memories

STT-MRAM: Unlimited endurance like DRAM, but with much lower power consumption in
standby.

A. D. Kent and D. C. Worledge, Nature Nano 10 (2015). 5

(STT-MRAM) Spin torque transfer magnetic RAM

6
 What do we want for STT-RAM?
“switching
current”

“damping” thickness of memory layer

Low switching currents  small alpha, small volume
area of memory layer
thermal “attempt” time (~ 1 ns)

“decay” time of memory state

High stability  large anisotropy (Hk)

For >10 year stability, need For scalability:

7
 Ferromagnetic resonance in a nutshell
The Gilbert equation: The magnetic analog to Ohm’s law

Reactive: “Larmor” Lossy: “damping”

Resonance Linewidth 8
 Overview

• STT-MRAM: Background and motivation

• VNA-FMR: Measuring damping and anisotropy in

blanket films.

• H-MOMM: Measuring damping and anisotropy in e-

beam patterned structures.

9
 Instrumentation: VNA-FMR
•10 MHz < f < 67 GHz
Vector network •Maximum field: 3 T
analyzer •Coplanar waveguides with 50
μm wide center conductor
•1 Watt max microwave power

Port 1 magnetic Port 2

sample

50 coplanar
waveguide
“Perpendicular
FMR”

10
Ferromagnetic Resonance (FMR) @ NIST Boulder

3-Axis Superconducting Magnet Perpendicular Geometry

(3 Tesla) (2.4 Tesla)

11
 NIST coplanar waveguides

• “Stitching” to electrically connect all three

ground planes.

• Prevents mode hybridization for f < 40 GHz.

12
 Ferromagnetic Resonance (FMR)

Perpendicular geometry Vector Network Analyzer FMR (VNA-FMR)

• 67 GHz bandwidth, 3 T perpendicular fields
1.5
Imaginary S21 (arb. units)

15 GHz 25 GHz 22 GHz

0.5 15 GHz

Real S21 (arb. units)

18 GHz 22 GHz 18 GHz
1.0
25 GHz
Ex: NIST-grown 0.0
0.5
CoFe/Ni
-0.5
multilayers
0.0
Data -1.0 Data
tCoFe = 0.48 nm Fit tCoFe = 0.48 nm Fit
-0.5
0.9 1.0 1.1 1.2 1.3 1.4 0.9 1.0 1.1 1.2 1.3 1.4
Magnetic Field, µ0H (T) Magnetic Field, µ0H (T)

13
 Extracting anisotropy and damping from FMR

Slope  Damping, α
Slope  g-factor
y-intercept  Inhomogeneous
y-intercept  Effective magnetization, Meff
linewidth broadening ΔH0
65
nm
Resonance Field, µ0Hres (T)

2.0 Data 8 nm 60 Data

12
0.5 .
Fit = 55 Fit =0

Linewidth, µ0∆H (mT)

t CoFe m 50 t CoFe
8n
1.5 0.4 45 8 nm
= 40 = 0.3
t CoFe Ex: NIST-grown t CoFe
35
CoFe/Ni multilayers
1.0 nm 30 m
.38 0.48 n
=0 25 t CoFe
=
t CoFe 20
15 m
0.5
.12n
m
= 0.58 n
0 10 t CoF
e
=
t Co
Fe 5
0.0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Frequency (GHz) Frequency (GHz) 14
Damping parameter in Co90Fe10/Ni multilayers

0.035
Damping Parameter, α Multilayers
0.030 6.3 nm
10 nm
0.025 untextured
Alloys
0.020 6.3 nm
10 nm

0.015

0.010
6 nm
0.005 10 nm (A typical
bulk
0.000
value…)
0 10 20 30 40 50 60 70 80 90 100
at. % Co90Fe10

J.M. Shaw, APL, 99, 012503 (2011)

15
 Damping for “conventional” Ta/CoFeB/MgO

!!!!

• Small thickness required for high anisotropy (“interfacial anisotropy”).

• Small thickness results in higher damping, but with quadratic
dependence on reciprocal thickness. Interfacial?!?

16
 MgO “sandwiches”

IEEE, International Electron Devices Meeting, 677 – 680 (2012)

“Storage layer”
MgO “spacers”

17
 Intel CoFeB sandwich material
MgO(2 nm) / Co0.6Fe0.2B0.2(dm) / Ta(0.4 nm) / Co0.6Fe0.2B0.2(dm) / MgO(2 nm)

dm = 1.2 nm

dm = 0.8 nm

2.6x smaller than for single-layer

CoFeB with the same anisotropy!

18
 Overview

• STT-MRAM: Background and motivation

• VNA-FMR: Measuring damping and anisotropy in

blanket films.

• H-MOMM: Measuring damping and anisotropy in e-

beam patterned structures.

19
 Damping and finite size effects

“Storage layer”
Question: Is the damping
measured with an unpatterned
film representative of damping in
structures smaller than 30 nm?
20
 Finite size effects: “Drum-head” eigenmodes

http://www.fas.harvard.edu/~scidemos/OscillationsWaves/VibratingDrumhe
ad/VibratingDrumhead.html

http://www.ima.org.uk/viewItem.cfm-cit_id=383396.html 21
 Nanomagnet eigenmodes: Micromagnetics
160 nm x 350 nm x 5 nm Permalloy in zero field:

100 nm x 120 nm x 10 nm Permalloy in zero field: McMichael and Stiles, JAP 2005
“Center modes”

“End modes” J. Shaw, et al. PRB 2009

22
 Advantage: MOKE
y ψk
x
E E

“Magneto-optic Kerr effect” (MOKE)

• Non-invasive.
• Local probe. (Diffraction limited ~ 500 nm).
• Vector sensitive.
• Broadband/high speed compatible.

John Kerr (1824-1907)

Michael Faraday (1791 –1867)
23
 Challenge of measuring small magnets

Patterned magnet

Optical spot
Increased “resolution”, but at
the expense of sensitivity…

Signal (a.u.) 1.0

0.8

0.6

If sensitivity gap can be 0.4

overcome, an example 0.2 Conventional MOKE

of “device-defined 0.0
0 250 500 750 1000
resolution” Diameter (nm)

24
 Prior art: fs pump-probe
Holger Schmidt, UC Santa Cruz (Time-resolved MOKE)

Ni nanomagnet
500 nm diameter

Ni nanomagnet
150 nm diameter

Z. Liu, et al., APL 98, 052502 (2011)

25
 H-MOMM technique summary
“MOKE angle”
Detection:
Er
sample Etotal detector

Ek Ek
“analyzer” ELO
(polarizer) “local oscillator”

Excitation:
Ni80Fe20 Al2O3

RF field
CPW
100 nm

z
x y Hb bias field
26
 H-MOMM Advantage: SNR
Signal-to-noise estimate (ONLY shot noise and detector noise.)

27
 H-MOMM diagram
Polar MOKE
Two single-frequency CW lasers RF field
electro- CPW
magnet
sample

probe z
pol x y Hb bias field

PD1

BS2
+
PD2
BS1
-
dc signal out
LO

Differential heterodyne detection

Heterodyne mixing as microwave source

28
original Experiment background

Single nanomagnets
Nominal sizes: 50, 100, 200, 400 nm
4 nm Si3N4
10 nm Ni80Fe20
3 nm Ta
Sapphire

29
 H-MOMM measured spectra
Spectra of a 100 nm Permalloy (Ni80Fe20)
nanomagnet
with t=10 nm
7 16.5 GHz

6
Amplitude (a.u.)

5
4
13.1 GHz
3 Ni80Fe20 Al2O3
2
9.9 GHz
1

50 100 150 200 250 300 350

µ0H (mT)
100 nm
30
Damping examples

15

µ0∆H (mT)
10

5
Center
200 nm
0

15

µ0∆H (mT)
10

5
End
200 nm
0
4 8 12 16
f (GHz)

31
 Damping vs. size/mode
H-MOMM Data

0.011
Damping Parameter, α

nd Mode Damping has nontrivial

E
0.010
dependence on both
nanomagnet size AND
0.009
eigenmode profile.
Center Mode
0.008
NiFe film
0.007
100 200 300 400
Nanomagnet Width (nm)

Nembach, Shaw, Boone, & Silva PRL 2013

32
 Nonlocal damping theory
Phenomenological damping in metals (Bar’yakhtar JETP 1984)

Exchange-mediated damping

Mode curvature affects damping! Larger curvature = larger damping

Transverse intralayer spin diffusion theory (Tserkovnyak, Hankiewicz, Vignale PRB 2009):

33
 Damping vs. size/mode
Simulation w/ nonlocal damping
H-MOMM Data

Damping Parameter, α
0.011 0.011
Damping Parameter, α

0.010
End Mode
0.010 End Mode

0.009 0.009
Center
Center Mode 0.008 Mode
0.008
NiFe film
0.007
0.007 100 200 300 400
100 200 300 400 Nanomagnet Width (nm)
Nanomagnet Width (nm)

~ 100x too small to explain

Tserkovnyak, et al.:
our data!
Nembach, Shaw, Boone, & Silva PRL 2013
34
 New H-MOMM data
Damping for end-modes vs. size and film thickness
0.014

0.012

α
0.009
t=3 nm
End-Mode t=10 nm
t=15 nm
0.007

100 200 300 400

nanomagnet size (nm)

(H. Nembach, et al., in preparation)

35
 Summary
• STT-MRAM: An promising memory for low-power applications.

• Need to characterize damping and stability in advanced materials.

• Blanket thin films: VNA-FMR.

Ex: Low damping in engineered sandwich memory layers.

• Patterned structures: H-MOMM.

Ex: Curvature-dependent “non-local” damping with eigenmodes.

T. J. Silva, et al., “Characterization of Magnetic Nanostructures for Spin-Torque

Memory Applications with Macro- and Micro-Scale Ferromagnetic Resonance,” in
Characterization and Metrology for Nanoelectronics, eds. Zhiyong Ma and David
Seiler, (to be published by the end of this year by Pan Stanford Publishing.)
36