Deprocessing and defect analysis of GaN/AlGaN HEMTs. Patrick Whiting, Ray Holzworth Dr. Nicholas Rudawski, and Dr. Kevin Jones

Transcription

1 Deprocessing and defect analysis of GaN/AlGaN HEMTs Patrick Whiting, Ray Holzworth Dr. Nicholas Rudawski, and Dr. Kevin Jones

2 Scientific Approach FLOORS TEM SEM SEM LEAP t=0, As Built TEM OSTS Gate Metal Reaction? t>0, Degradation

3 Motivation G Source DF STEM Drain 10nm DF STEM Defect density varies along the length of HEMT devices. Formation is a stochastic process. 200nm

4 Motivation G 150um Gate / 5um per TEM crosssection = 30 TEM cross-sections 40 TEM cross-sections * 1 hr per section = 30 hours of FIB time = $2,100 per device Source Drain 40 TEM cross-sections*.5 hr per section = 15 hours of TEM time = $1,050 per device One solution is to determine defective quality by looking at lots of cross-sections. This is expensive!

5 Motivation G SEM Source Drain 100nm DF STEM Another option is to use the Slice and View method to reduce the number of TEM crosssections. Resolution suffers, however. 100nm

6 Motivation G *Makaram, et. al., Applied Physics Letters 96, 23 (2010) etch pit formation Source Drain The simplest (and arguably, the best) option is to deprocess the sample and perform PlanView SEM.*

7 Deprocessing: Initial Conditions SiN Au GaN 5um Drain AlGaN Gate Source Ni

8 Deprocessing: Nitride Strip AlGaN GaN 5um Drain Source Ni Au SiN Gate 6:1 HF:NH3F Bath for 9 minutes

9 Au Ni Drain SiN Source Deprocessing: Gold Precipitate Clean AlGaN GaN 2um TFAC (FeCN/KCN solution) bath for 20 minutes TFAC Gold Etchant produced by Transene Corp.

10 Deprocessing: Second Nitride Strip AlGaN Drain Ni Source Au Gate GaN 2um 6:1 HF:NH3F Bath for 6 minutes

11 Ni Drain Source Deprocessing: Metal Strip AlGaN GaN 2um TFAC (FeCN/KCN solution) bath for 3 hours TFAC Gold Etchant produced by Transene Corp.

12 Ni AlGaN GaN 2um 6:1 HF:NH3F Bath for 3 minutes Ultrasonic Methanol bath for 20 minutes Drain Source Deprocessing: Ultrasonic Clean

13 Drain Source Comparison: delalamo Group (MIT) 2um Left: BOE Nitride Strip with Aqua Regia Metal Etch (MIT) Right: BOE Nitride Strip with FeCN/KCN (UF)

14 Selectivity The process does not appear to attack AlGaN Platinum Nickel ~13nm AlGaN ~13nm AlGaN GaN 10nm Control 10nm GaN Post-Etch Sample

15 Analysis of AFRL1813 Presence of reaction-based defect? 10nm AFRL1813: DF STEM 200nm 1.77 um from Drain AFRL1813Plan-View SEM Gate Edge ~60nm 200nm AFRL1813 Plan-View SEM

16 Conclusions Slice-and-View SEM and TEM analysis may not be the best way to perform statistical analysis of gate electrode defects. A new technique for deprocessing the channel regions of stressed devices has been developed. This new technique was used in the analysis of a previously stressed device. SEM analysis indicates that the technique shows promise as a means of viewing defects in plan-view.

17 Future Work Development of complementary AFM technique for characterization of the denuded channel regions. Combination of the deprocessing technique outlined above with AFM/SEM to investigate HEMT structures biasstressed for longer time periods.

18 A System for the Characterization of Electronic Defects in AlGaN/GaN HEMTs Patrick Whiting, David Cheney and Erica Douglas Dr. Brent Gila, Dr. Kevin Jones, Dr. Stephen Pearton

19 Where can traps come from? Reactive Gate AlGaN 2DEG GaN Some trap states arise from physical defects which are easy to identify via TEM, SEM, optical microscopy, etc.

20 Where can traps come from? Reactive Gate Interface States AlGaN 2DEG GaN Point Defects Threading Dislocations Other trap states arise from defects which are so small that they can't be seen except electrically. Nevertheless, both varieties of traps are important.

21 Free/Trapped Carrier Dynamics EC Band to Band Excitation EG=hc/λ ET ΔE=hc/λ Carrier Excitation from a trap state EV

22 Free/Trapped Carrier Dynamics EC Electron Capture =σvthn Electron Emission =σvthncexp(-δe/kt) ET Hole Capture =σvthp Hole Emission =σvthnvexp(-δe/kt) EV

23 Existing Methods: PL, CL, PCS Advantages PL CL PCS Direct, light-based determination of trap densities with respect to energy. Variable temperature is not required in order to determine the energy of trap levels. Disadvantages Steady state measurement: no data is obtained with regards to rate of decay. Carrier dynamics must be optically active to be observed.

24 Existing Methods: PL, CL, PCS Advantages Direct measurement of decay rates associated with carrier capture and emission of trapped carriers. Disadvantages *Polyakov et. al., J. Vac. Sci. Technol. B 18.3 (2000). Temperature scans are required in order for the technique to measure trap energy. Temperature scans make the resolution of specific energies very difficult. It is also difficult to reproduce results.

25 Thermal Variance Method (DLTS) ET1 Counts Current UV Diode On time EC ET2 ET1 EV ET2 UV Diode On: Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: Trapped Carrier Emission Temperature = 200K

26 Thermal Variance Method (DLTS) ET1 ET2 Counts Current UV Diode On time EC ET2 ET1 EV UV Diode On: Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: Trapped Carrier Emission Temperature = 350K

27 Thermal Variance Method (DLTS) Counts Current UV Diode On time EC ET2 ET1 UV Diode On: Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: EV Trapped Carrier Emission Temperature = 200K

28 Thermal Variance Method (DLTS) Non-Linear Distortion Counts Current UV Diode On time EC ET2 ET1 UV Diode On: Band to Band Excitation Free Carrier Capture UV Diode Off: EV Trapped Carrier Emission Temperature = 350K

29 Why DLTS won't work well for HEMTs Reactive Gate Interface States AlGaN 2DEG GaN Point Defects Threading Dislocations There are lots of defects which will yield an energy band of trap states within the forbidden gap!

30 The Solution: Take the Best of Both Worlds Hg Ar cl am p Device Under Test Steady-State Biasing Current Transient Analysis / Control Computer Filter Settings Pulsing Conditions d e ls u P U V i D e d o

31 Optical Scan Transient Spectroscopy (New!) ET1 Counts Current UV Diode On time EC ET2 ET1 EV ET2 UV Diode On: Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: Trapped Carrier Emission Secondary Light Source Off

32 Optical Scan Transient Spectroscopy (NEW!) ET1 Counts Current UV Diode On X time ET2 EC UV Diode On: ET2 ET1 Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: EV Emission from ET1 only Secondary Light = ET2

33 Optical Scan Transient Spectroscopy (NEW!) ET1 Counts Current UV Diode On XX time ET2 EC UV Diode On: ET2 ET1 Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: EV No Emission. Secondary Light = ET1

34 Optical Scan Transient Spectroscopy (NEW!) Counts Current UV Diode On time EC ET2 ET1 UV Diode On: Decay Rate Band to Band Excitation Free Carrier Capture UV Diode Off: EV Trapped Carrier Emission Secondary Light Source Off

35 Optical Pulsing Method (OSTS) Linear Superposition Counts Current UV Diode On X time EC ET2 ET1 UV Diode On: Band to Band Excitation Free Carrier Capture UV Diode Off: EV ET1, Band edge emission Secondary Light = ET2

36 Example Experiment Emission Peak? Current Arc Lamp On time 360nm 400nm 440nm 0.08 EC 540nm 570nm 650nm ET IDS (A) On-State analysis EV B A Time (s)

37 System Design: Minimum Decay Rates Data collection speed and rise/fall times for switching determine minimum decay rate. 10us collection time for analog electronics which interface with the control computer 10ns rise/fall time for the pulsed UV diode. Both shallow levels and deep levels can be resolved by OSTS.

38 System Design: Decay Rate Resolution Algorithm choice is the 1 determining factor in resolving two similar rates 0.4 of decay di/d(log[t]) vs log[t] is simple but has very poor resolution. Resolution can be improved by squaring the differential. Fast Laplace Transforms greatly improve decay *Tapajna et. al., Applied Physics Letters 97, (2011) rate resolution. FWHM (log[s]) 1.2 Exponential Power

39 A Comparison* Figure of Merit OSTS DLTS Minimum Decay Rate 50 us 1ms Decay Rate Resolution <1 Order of Magnitude >1 Order of Magnitude Trap Energy Determination Optical Thermal Trap Positioning Optical Excitation Variable Biasing *Tapajna et. al., Applied Physics Letters 97, (2011)

40 Conclusion A new method for observing trapping effects in AlGaN/GaN HEMTs, tentatively named Optical Scanning Transient Spectroscopy has been proposed. The OSTS system currently being developed at University of Florida has been described and an initial pulsed light experiment has been presented. Future work will include integration of a UV Diode into the OSTS system for UV excitation experiments, integration of a Fast Laplace Transform for improved spectral resolution and an initial proof of concept experiment.

41 Yearly Summary A new method for deprocesing AlGaN/GaN HEMTs has been developed. This new method shows promise as a means of sample preparation for plan-view imaging either via SEM or AFM. Work continues in building a system capable of detecting and analyzing the transient signal of deep level traps in HEMT structures. A new method for measuring these deep traps, Optical Scanning Transient Spectroscopy, has been proposed. Future work on stressed devices will focus both on further improvements to the deprocessing technique described, including integration with AFM, as well as the further development and experimental validation of OSTS.