Realism in Nanosensing: Hard-Won Insights from the Trenches

Realism in nanosensing: hard-won insights from the trenches

Michael Roukes Robert M. Abbey Professor of Physics, Applied Physics, & Bioengineering Co-Director, Kavli Nanoscience Institute California Institute of Technology [email protected] http://nano.caltech.edu Abstract: Nanosensors have the unprecedented capability to interact with absolute numbers of analytes down to the single molecular scale. The susceptibility of nanosensors to external perturbation, given their size, is unprecedented -- and this essential physics offers exceptional new opportunities. Indeed, we are here to explore, harness, and even celebrate these attributes! As a archetypal case study of nanosensor physics, I will focus upon resonant NEMS sensors. I will describe the immense increase in responsivity obtained with device size reduction that yields, for example, atomic mass resolution. As an archetypal case study of nanosensor physics, we’ll focus upon resonant NEMS sensors. I will describe the immense increase in responsivity obtained with device size reduction that yields, for example, atomic mass resolution. We will explore several noise mechanisms and backaction effects imposing ultimate physical limits to the sensitivity, and how matching considerations make it extremely challenging to embed these devices in practical measurement systems. The unique properties of nanodevices, however, are not a panacea that absolves us from the need to contend with long-standing, overarching challenges posed by real-world sensing applications. From about a decade and a half’s work focused on metamorphosing nanosensors into nanosystems through large-scale integration, nature has taught us some hard lessons. In my talk I will share several of these, using examples from NEMS and other sensor systems. Among them are: (a) any stand-alone sensor is unlikely to be up to the full dynamic range of challenges posed in the real world; (b) complex combinations of real-world analytes are unlikely to be resolved by any “one-stop” generic approach; (c) arrays of sensors can often give enhanced performance over individual devices, but achieving such gains may require exceptionally careful attention to device variances; (d) careful choices of platform materials are essential to insure a rapid transitioning into large-scale integration and production en masse. Although these sobering considerations may potentially force an initial downselection from amongst the panoply of exciting possibilities of emerging nanosensors; learning from the past, and considering real- world (not academic laboratory) conditions, will accelerate realization of practical nanosensor systems. 22 March 2012 ML Roukes © Caltech 2012 2 single-molecule NEMS-MS AK Naik, MS Hanay, WK Hiebert, XL Feng, and ML Roukes, Nature Nanotechnology 4, 445-449 (2009)

0 background fluctuations -5 BSA -10 s ingle m

(kHz) o lecule -15 adsor p(precipitoustion e jumps) vents

-20

-25

-30

Frequency Shift Frequency z -35 y x -40 0 100 200 300 400 500 600 700 800 900 Time (s)

NAIK, Akshay HANAY, Selim HIEBERT, Wayne M.L. Roukes Staff Physicist3 © Caltech/CEA-LETIGrad Student 2011 Postdoc 22 March 2012 the promise Nanosensors have the unprecedented capability to interact with absolute numbers of analytes down to the single molecular scale. The susceptibility of nanosensors to external perturbation, given their size, is unprecedented -- and this essential physics offers exceptional new opportunities. Indeed, we are here to explore, harness, and even celebrate these attributes!

22 March 2012 ML Roukes © Caltech 2012 4 responsivity, noise, backaction, mismatch

22 March 2012 ML Roukes © Caltech 2012 5 physics: responsivity, noise, backaction, mismatch As an archetypal case study of nanosensor physics, we’ll focus upon resonant NEMS sensors. I will describe the immense increase in responsivity obtained with device size reduction that yields, for example, atomic mass resolution. We will explore several noise mechanisms and backaction effects imposing ultimate physical limits to the sensitivity, and how matching considerations make it extremely challenging to embed these devices in practical measurement systems.

22 March 2012 ML Roukes © Caltech 2012 6 frequency-shift based mass sensing

frequency shift

Δ f ~ − ( f0 /2M eff )⋅δ m

“gain” added mass responsivity mass

resonant response z

y x doubly-clamped beam resonator

M.L. Roukes 7 © Caltech/CEA-LETI 2011 22 March 2012 minimum resolvable mass is determined by “gain” and noise

∂Meff -1 δM ≈ δω0 ~ R σ A(τ ) mass ∂ω0 Allan a measure of resolution Deviation frequency-fluctuation noise

Mass Responsivity

M.L. Roukes 8 © Caltech/CEA-LETI 2011 22 March 2012 Maximizing mass responsivity leads to minimizing the resolvable mass

∂Meff -1 δM ≈ δω0 ~ R σ A(τ ) mass ∂ω0 Allan a measure of resolution Deviation frequency-fluctuation noise

large mass responsivity Mass Responsivity For any complex mechanical mode, modeled as a damped simple harmonic oscillator: ω R ≈− 0 2Meff 1 ∝ huge benefit in 4 scaling downward !

M.L. Roukes 9 © Caltech/CEA-LETI 2011 22 March 2012 size-scaling the minimum resolvable mass: resonant inertial mass sensors

quartz crystal microbalances

M eff ~1−100mg; δ m ~1μg

SAW/FPW sensors (MEMS)

M eff ~1μg −1mg; δ m~1ng

microcantilevers (MEMS)

M eff ~1ng−1μg; δ m ~1fg

NEMS

M eff ~1fg−1pg; δ m ≤ 1zg

M.L. Roukes 10© Caltech/CEA-LETI 2011 22 March 2012 Massive increase in responsivity: zeptogram mass sensitivity ( 1zg = 10-21g )

0 SiC doubly clamped beam NEMS UHF bridge configuration -200 ( f0 ~ 133 and 190 MHz ) Ultralow noise PLL readout -400 ~100 zg Nano Lett. 6, 583 (2006) -600 NEMS shutter -800 0 Frequency Shift (Hz) Shift Frequency -1000 nozzle -500

0 50 100 150 200 250 300 350 -1000 Time (sec)

-1500

-2000 133 MHz Mass

Frequency Shift (Hz) responsivity -2500 Ya-Tang Yang, Carlo Callegari , 190 MHz ~1Hz/zg Xiaoli Feng, Kamil Ekinci, MLR -3000 0 1000 2000 3000 4000

M.L. Roukes Mass (zeptograms)© Caltech/CEA-LETI 2011 22 March 2012 11 BUT… minimizing frequency-fluctuation noise is also critical !!! M ∂ eff -1 δ M ≈ δω 0 ~ R σ A (τ ) mass ∂ω 0 Mass Allan a measure of resolution Responsivity Deviation frequency-fluctuation noise

frequency resolution optimized phase-locked loop frequency tracking

noise processes (f) a log S

012

Frequency f /fc

M.L. Roukes 12© Caltech/CEA-LETI 2011 22 March 2012 Real-time zeptogram mass sensitivity ( 1zg = 10-21g )

0 SiC doubly clamped beam NEMS UHF bridge configuration -200 ( f0 ~ 133 and 190 MHz ) Ultralow noise PLL readout -400 ~100 zg Nano Lett. 6, 583 (2006) -600 NEMS shutter -800 100 0 Frequency Shift (Hz) Shift Frequency -1000 nozzle 10 Mass

-500 (zg) resolution

m ~7 zg 0 50 100 150 200 250 300 350 δ 1 -1000 Equivalent to about Time (sec) 0.1 0 2000 4000

30 Xe-1500 atoms… Time (s)

-2000 133 MHz Mass

Frequency Shift (Hz) responsivity -2500 Ya-Tang Yang, Carlo Callegari , 190 MHz ~1Hz/zg Xiaoli Feng, Kamil Ekinci, MLR -3000 0 1000 2000 3000 4000

M.L. Roukes Mass (zeptograms)13© Caltech/CEA-LETI 2011 22 March 2012 physics: responsivity, noise, backaction, mismatch As an archetypal case study of nanosensor physics, we’ll focus upon resonant NEMS sensors. I will describe the immense increase in responsivity obtained with device size reduction that yields, for example, atomic mass resolution. We will explore several noise mechanisms and backaction effects imposing ultimate physical limits to the sensitivity, and how matching considerations make it extremely challenging to embed these devices in practical measurement systems.

22 March 2012 ML Roukes © Caltech 2012 14 sources of frequency-ﬂuctuation noise

22 March 2012 ML Roukes © Caltech 2012 15 responsivity, noise, backaction, mismatch # " !

$%" &' !

22 March 2012 ML Roukes © Caltech 2012 16 physics: responsivity, noise, backaction, mismatch As an archetypal case study of nanosensor physics, we’ll focus upon resonant NEMS sensors. I will describe the immense increase in responsivity obtained with device size reduction that yields, for example, atomic mass resolution. We will explore several noise mechanisms and backaction effects imposing ultimate physical limits to the sensitivity, and how matching considerations make it extremely challenging to embed these devices in practical measurement systems.

22 March 2012 ML Roukes © Caltech 2012 17 Cantilever tales: “Can a gauge factor ever be “too big”? Scaling down: semiconducting nanopiezoresistor issues

RF electronics

signal Pros: 50Ω >>1MΩ 1. A large gauge factor V. serious impedance Cons: matching problem at device RF & VHF and beyond 1. V. high resistance 2. Scaling exacerbates issues 3. Noisy (Johnson + 1/f) 4. Large temp. coefficients 5. Difficult to fabricate (epi Si) 6. Expensive to produce

19 Technology: self-sensing NEMS resonators

10μm 2μm 500nm 200nm

Piezoresistively-detected resonant response from a family of SiC nanocantilevers to a 1nN AC drive signal vs. frequency using metallic piezoresistors, at room temperature in vacuo.

Mo Li, HX Tang, ML Roukes, M.L. Roukes Nature Nanotechnology 2, 114 (2007) 20 © Caltech 2010-2012 22 March 2012 Technology: self-sensing NEMS resonators

10μm 2μm 500nm 200nm

3.9×10-15 m/√Hz

Piezoresistively-detected resonant response from a family of SiC nanocantilevers to a 1nN Thermomechanical noise spectrum and Lorentzian fit AC drive signal vs. frequency using metallic piezoresistors, at room temperature in vacuo. (red trace) for a 123 MHz nanocantilever measured at room temperature and atmospheric pressure. Mo Li, HX Tang, ML Roukes, M.L. Roukes Nature Nanotechnology 2, 114 (2007) 21 © Caltech 2010-2012 22 March 2012 responsivity, noise, backaction, mismatch

" # ! & $' % $

22 March 2012 ML Roukes © Caltech 2012 22 Systems: ingredients

# ( ' ( &

) " * +& I

V )

!& ) ) %*+ * +

22 March 2012 ML Roukes © Caltech 2012 23 Resonant Systems: Generic Platforms

# ( ' ( &

) " * +& I

V )

!& ) ) %*+ * +

22 March 2012 ML Roukes © Caltech 2012 24 Resonant Systems: Generic Platforms

""% - 0 0 & ! 2 !3 • (" )!!#&"(" "% #! • "!")1!"" #!)" !)!"2) '1!!" %" )3 • '1!1!!" ")!'1&"0"% - &!! &#!!! • '& ! +# !!240#3!'%!!#! . %!0 '& +" "" !" !. ! "+ ! "+ 2 !! !3 '&0 0# ! "+ ! , "22 March 2012 ML Roukes © Caltech 2012 ,/".25 generalization I will attempt to generalize the discussion (a little) by considering other state-of-the-art sensors such as resonant photonic devices.

22 March 2012 ML Roukes © Caltech 2012 26 Photonic micro-ring resonators

$' $ # $ $%*())&

$# !%*()(& # $%*((+&

22 March 2012 ML Roukes © Caltech 2012 27 Zero-power sensors: a promising idea? One last organizers’ thought to address: is there a way, power- wise, to get your data for nothing (and your chips for free?) Let’s take the archetypal case of thermoelectric µﬂuidic calorimeters.

! " $% • ! • $ • % • ! *& • !( & ') •

22 March 2012 ML Roukes © Caltech 2012 28 (hard) lessons learned The unique properties of nanodevices, however, are not a panacea that absolves us from the need to contend with long-standing, overarching challenges posed by real-world sensing applications. From about a decade and a half’s work focused on metamorphosing nanosensors into nanosystems through large- scale integration, nature has taught us some hard lessons.

22 March 2012 ML Roukes © Caltech 2012 29 (hard) lessons learned Among them are: (a) much, if not most of the work in applying nanosensors must be invested in embedding them into functional systems;

22 March 2012 ML Roukes © Caltech 2012 30 From proteins in solution to vacuum-based adsorption

• electrospray Injection

• dual differential 760Torr pumping ESI Needle (b) capillary • dual hexapole 1Torr ion gudes

• faraday cup 10mTorr Top • phase-locked hexapole 10-6 Torr (c) UHF NEMS 2.5 mass sensor Bottom hexapole -65 (~428 MHz) 2.0 (dB) -70 Phase

Sample Stage 1.5 <10-8 Torr -75 Wayne Hiebert (T= 40K) 1.0 Selim Hanay -80 Magnitude Akshay Naik Magnet Philip Feng -85 0.5 428.0 429.0 Steve Stryker (a) (d) Frequency (MHz) (2004-08)

M.L. Roukes 31© Caltech/CEA-LETI 2011 22 March 2012 (hard) lessons learned (b) careful choices of platform materials are essential to insure a rapid transition into large-scale integration and production en- masse.

22 March 2012 ML Roukes © Caltech 2012 32 The Alliance for Nanosystems VLSI

Caltech – Kavli Nanoscience Institute, Pasadena CEA/LETI-MINATEC, Grenoble

Functional validation of nanosystems Very-large-scale Integration • >20 years in nanosystems and nanosciences • 200 & 300 mm technological facilities • understanding of underlying physics: ultimate limits • Systems architecture and integration • NEMS (nanoelectromechanical systems) (from design to packaging) • Physics, Biophysics, Engineering • Applied research / applications • Quick prototyping; Kavli Nanoscience Institute • Technology transfer to industry

22 March 2012 33 $B-scale microelectronic foundry required…

200mm • $B-scale “foundry” • fabrication of millions of nanodevices via standard 200mm (8”) VLSI processes

CEA/LETI-MINATEC, Grenoble & Kavli Nanoscience Institute, Caltech

M.L. Roukes 34© Caltech/CEA-LETI 2011 22 March 2012 $B-scale microelectronic foundry required…

200mm

Progressive magnifications

M.L. Roukes 35© Caltech/CEA-LETI 2011 22 March 2012 (hard) lessons learned (c) any stand-alone sensor is unlikely to be up to the full dynamic range of challenges posed in the real world;

22 March 2012 ML Roukes © Caltech 2012 36

Microsystems for Microsystems gas phase chemical analysis phase chemical gas The “electronic nose” paradigm: a chemical sensor array

Gaseous Sensor Analyte Array

Chemical “fingerprint”

l a n g i s

sensor

M.L. Roukes 38 © Caltech 2010-2012 22 March 2012 The “electronic nose” paradigm: a chemical sensor array

Gaseous Sensor Analyte Array

Chemical “fingerprint”

l a n g i s

sensor Well-documented problems with the e-nose paradigm: - crosstalk: “fingerprint” changes with exposure to multiple analytes - environmental fluctuations: e.g. temperature, humidity - saturation: hard to look for analytes at few ppt level (“t”=trillion) in the presence of few percent H20, 1000ppm smog, etc.

M.L. Roukes 39 © Caltech 2010-2012 22 March 2012 Caltech-Sandia Project: microGC + NEMS sensor array

species species … species group A group B group M

microfluidicN E M S N sample o s e : Microscale column (Sandia) preconditioning and chemically-functionalized multiplexed,

NEMS arrays

Chemisorb-Coated NEMS Resonator Arrays (Caltech)

M.L. Roukes 40 © Caltech 2010-2012 22 March 2012 Chemical functionalization preferentially “sorbs” analytes from ambient (1 Atm)

Ingredients: Chemisorptive NEMS Sensor

(a) chemisorptive coating

μGC pulse (b) resonant NEMS mass sensor 1ppt CWA to (c) control-loop readout MGA yields a pulse from GC that is ~10ppb for 10ms (typ.)

Areal responsivity: ~ few pg cm-2Hz-1

Frequency Resolution: ~1×10-8

M.L. Roukes 41 © Caltech 2010-2012 22 March 2012 Ultra-sensitive NEMS-based cantilevers for sensing, scanned probe and very high-frequency applications Mo Li, HX Tang, ML Roukes, NEMS mass sensing at 1 Atm Nature Nanotechnology 2, 114 (2007)

Real time chemisorption measurements with two PMMA-coated NEMS resonators. Displayed are frequency shift steps upon of adsorption of difluoroethane gas molecules onto device 8 MHz surface.

These measurements are carried out at atmospheric pressure & room temperature.

The top and bottom traces are measured with 128 MHz 8 MHz and 128 MHz nanocantilever devices, respectively.

The minimum resolvable mass is ~100 zg. M.L. Roukes 42 © Caltech 2010-2012 22 March 2012 Nanoelectromechanical Resonator Arrays for Ultrafast, Analytical grade performance achieved, from devices Gas-Phase Chromatographic Chemical Analysis Mo Li, EB Myers, HX Tang, SJ Aldridge, HC McCaig, millions of times smaller than anything before JJ Whiting, RJ Simonson, NS Lewis, ML Roukes, Nano Letters 10, 3899 (2010) • Frequency shift response to DIMP (a CWA simulant) is linear from ppm concentrations -> sub ppb levels • Sub-ppb sensitivity is comparable to that attained with state-of-the-art laboratory technologies for ultralight (non-sticky) analytes

RMS noise ~ 300 zg DIMP sensitivity (1 zg = 10-21 g) 200 part-per-trillion concentration sensitivity attained at 1 Atm

M.L. Roukes 43 © Caltech 2010-2012 22 March 2012 Miniscule NEMS footprint: state-of-the-art sensing in chip-scale GC columns

M.L. Roukes 44 © Caltech 2010-2012 22 March 2012 Miniscule NEMS footprint: state-of-the-art sensing in chip-scale GC columns Sensor Array

Two devices with diff. selectivities: 1) DKAP: specifically sensitive to phosphonate nerve agents 2) PCL: highly absorptive, but unselective, polymer

2-sensor NEMS nose

unresolved cluster of analytes

Two devices with diff. selectivities: 1) DKAP: specifically sensitive to phosphonate nerve agents 2) PCL: highly absorptive, but unselective, polymer

M.L. Roukes 46 © Caltech 2010-2012 22 March 2012 (hard) lessons learned (c) complex combinations of real-world analytes are unlikely to be resolved by any “one-stop” generic approach;

22 March 2012 ML Roukes © Caltech 2012 47 Nanoelectromechanical Resonator Arrays for Ultrafast, Gas-Phase Chromatographic Chemical Analysis GC+“NEMS nose” provides superior performance Mo Li, EB Myers, HX Tang, SJ Aldridge, HC McCaig, JJ Whiting, RJ Simonson, NS Lewis, ML Roukes, Nano Letters 10, 3899 (2010) 2-sensor NEMS NEMS Nose nose + GC

differ ential + chemi reten cal affi tion tim nity e reso = en lution hanced disc M.L. Roukes rimin 48 © Caltech 2010-2012 ation 22 March 2012 Miniscule NEMS footprint allows for massive scale-up

# "

" %" &

Area: 2.5 mm x 250 m 25,000 devices

# "

" %" &

M.L. Roukes 50 © Caltech 2010-20122.5 μm 22 March 2012 (hard) lessons learned (d) arrays of sensors can often give enhanced performance over individual devices, but achieving such gains may require exceptionally careful attention to device variances;

l l 1 1 l ⎛ δ jk ⎞ R = ≈ δ ≈ R ⎜1+ ⎟ l tot ∑ m ∑ m 0 ⎜ ∑ ⎟ j 1 j 1 m j,k lm ⎝ ⎠ V ∑ ∑ 1 k R0, jk (1+ jk ) k R0 (1+ jk )

ΔRtot δ jk V = C V = C V 1 2 m out det bias det bias ∑ i Rtot j,k lm V 2 In terms of dissipated power per device: bias Pdevice = 2 . l R0 2⎛ δ δ jk ⎞ V C P R l ⎜ ⎟ out = det device 0 ⎜1+ ∑ ⎟ ⎝ j,k lm ⎠

jk δ jk sV = detC deviceP 0R∑ = detC deviceP totR∑ , where N = lm j,k m j,k N Best-case scenario: All N resonator signals add coherently: signal ~ N Noise adds incoherently: total noise ~ √N Arrays can yield up to √N enhancement of S/N M.L. Roukes 52 © Caltech 2010-2012 (~160x22 March 2012 for 25,000 devices!) Scale-up Nano Letters 12, 1269-1274 (2012) provides withimmense minimization SNR improvement of variance

(a) Top-view schematic of a 140 × 20 cantilever array. Individual devices are not visible in this image. The dotted red line shows the trajectory of the laser-based displacement transduction system used to acquire the spectra. (b) Optically detected spectra of from cantilevers within the array as a function of the position of the laser readout beam. (Beam width is approximately 10 M.L. Roukes 53 μm).© Caltech 2010-2012 22 March 2012

Scale-up Nano Letters 12, 1269-1274 (2012) provides withimmense minimization SNR improvement of variance

(a) Top-view schematic of a 140 × 20 cantilever array. Individual devices are not visible in this image. The dotted red line shows the trajectory of the laser-based displacement transduction system used to acquire the spectra. (b) Optically detected spectra of from cantilevers within the array as a function of the position of the Foundry attainable dimensional laser readout beam. tolerances below 1% provide access to (Beam width is approximately 10 majority of array-basedM.L. Roukes enhancement 54 μm).© Caltech 2010-2012 22 March 2012

(hard) lessons learned (c, revisited for ﬂuidic sensors) any stand- alone sensor is unlikely to be up to the full dynamic range of challenges posed in the real world;

Comparative advantages of mechanical biosensors JL Arlett, EB Myers, ML Roukes, Nature Nanotechnology 10, 3899 (2011)

56 Fluidic biosensors: a worthy goal – 1pM LOD/1m

Comparative advantages of mechanical biosensors JL Arlett, EB Myers, ML Roukes, Nature Nanotechnology 10, 3899 (2011)

57 But (non-speciﬁc) biological noise is a problem

-9 -1 Ka~10 M

Comparative advantages of mechanical biosensors JL Arlett, EB Myers, ML Roukes, Nature Nanotechnology 10, 3899 (2011)

58 Label-free capture probe afﬁnity is also a challenge

-11 -1 Ka~10 M X

Comparative advantages of mechanical biosensors JL Arlett, EB Myers, ML Roukes, Nature Nanotechnology 10, 3899 (2011)

59 …but, to end on a high note… Although these sobering considerations may potentially force an initial downselection from amongst the exciting possibilities from within the emerging panoply of nanosensors… …learning from the past, and considering real- world (not academic laboratory) conditions, will accelerate realization of practical nanosensor systems.